56 10 3494 https://omeka.ibu.edu.ba/files/original/06e97e59b1196ae4206d8a400a77b262.pdf 8d846e47c333719dbd78fd4d92de9253 Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource IT Senior Design Projects Description An account of the resource The IT Senior Design Projects (SDPs) category showcases innovative and practical final-year capstone projects developed by undergraduate and graduate students in the field of Information Technology. These projects represent the culmination of students' academic and technical expertise, demonstrating their ability to solve real-world problems through software and hardware solutions. Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Cinebh - Cinema Management System Author Author Sead Mašetić Abstract A summary of the resource. Cinema management involves handling reservations, scheduling screenings, and maintaining effective communication with customers. This project presents a Cinema Management System designed to automate these processes, reducing administrative workload and enhancing user experience. The system streamlines reservations, sends automated notifications for confirmations and reminders, and provides an intuitive interface for both customers and administrators. Developed using Spring Boot [6] for the backend and React [7] for the frontend, the system ensures scalability and maintainability. The development process followed a structured approach, starting with requirement analysis to identify challenges in manual cinema operations. The system was designed using a three-tier architecture, with a Spring Boot [6] backend handling authentication, reservations, and notifications, a React [7] frontend providing an intuitive interface, and a PostgreSQL [4] database for efficient data storage. Automated email and in-app notifications were integrated to keep users informed about their bookings. The Cinema Management System successfully automates key aspects of cinema operations, reducing manual workload and improving user experience. Testing confirmed the platform's efficiency, user-friendliness, and scalability, with responsive performance across different devices. By integrating automation and a modern technology stack, the system significantly improves operational efficiency for cinemas. Keywords Keywords. Cinema Management, Reservations, Real-Time Notifications, WebSockets, Venues Language A language of the resource English language cinema Management reservations software engineering https://omeka.ibu.edu.ba/files/original/405711553b574d8fdbc957a5deff9ac7.pdf 0c19e7ad5b81c69f372bca28387779a5 Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource IT Senior Design Projects Description An account of the resource The IT Senior Design Projects (SDPs) category showcases innovative and practical final-year capstone projects developed by undergraduate and graduate students in the field of Information Technology. These projects represent the culmination of students' academic and technical expertise, demonstrating their ability to solve real-world problems through software and hardware solutions. Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource CINEMA TICKET BOOKING - QUIKYTIX Author Author Adnan Brkić Abstract A summary of the resource. In a world where leisure time is precious, finding a hassle-free way to book cinema tickets can often be a challenge, leading to inefficiency and inconvenience. To address this issue, we introduce Quiky Tix, a user-centric platform designed to streamline the cinema ticket booking experience. Quiky Tix leverages modern web technologies such as Next.js, React, and TypeScript for robust development and utilizes Prisma and PostgreSQL for database management. The application offers a comprehensive movie list with titles, descriptions, and trailers, empowering users to make informed decisions. Through an intuitive interface, users can easily select seats and personalize their cinema experience. The implementation of Quiky Tix results in a seamless and efficient ticket booking process. Users can register effortlessly, accessing features like storing tickets with scannable barcodes. Timely notifications keep users informed about purchases and upcoming releases, enhancing their overall experience. Quiky Tix redefines the cinema ticket booking landscape, providing a user-friendly solution to inefficiencies in traditional methods. By combining ease of use with modern technology, the application revolutionizes the way people enjoy movies. Through its streamlined approach, Quiky Tix ensures a seamless and enjoyable movie experience for all users. booking management system https://omeka.ibu.edu.ba/files/original/526af53602e113658824c2078ccb6f9a.pdf 39f93afce0d809081027cb4a0ea905ed PDF Text Text Watkıns K. And Marsick V. (1997) Dimensions of The Learning Organization Questionnaire [survey] (Warwick, RI: Partners for the Learning Organization) Civil Law Notaries in Bosnia and Herzegovina: Actors in Preventive Justice Bakšić Šukrija1, Oruč Esad2 1University of Zenica, Faculty of Law, Zenica, Bosnia and Herzegovina, 2International Burch University, Sarajevo, Bosnia and Herzegovina E –mails: sukrijabaksic@gmail.com,eoruc@ibu.edu.ba Abstract Civil law notaries are professional lawyers and public officials appointed by the State to confer authenticity on legal deeds and contracts contained in documents drafted by them and to advise persons who call upon their services. Institution of the notary was introduced for the first time in the legal system of Bosnia and Herzegovina in 2007. Introduction of the office of notary was one of the steps taken to ensure independent and impartial judiciary and to adapt legal system with European Union law. Before its introducing there was no institution or legal profession which acted impartially on behalf of all parties to a contract or transaction. Notarial services are very wide and complex. It encompasses all judicial activities in noncontentious matters, ensure legal certainty to clients, thus averting disputes and litigation. As a guarantor of legal certainty, notary is one of the most important actors of preventive justice which include all means of reducing resort to the courts for the settlement of controversies. In this study we analyzed contribution of notary office to preventive justice in Bosnia and Herzegovina. Keywords: civil law notary, preventive justice, legal certainty, realising justice, avoiding disputes 412 �1.INTRODUCTION The 1995 General Framework Agreement for Peace in Bosnia and Herzegovina ended the 1992-95 war and created the independent state of Bosnia and Herzegovina (BiH). BiH consists of two entities and one district: The Federation of Bosnia and Herzegovina (FBiH) and the Republika Srpska (RS), as well as Brčko District of Bosnia and Herzegovina (BD BiH), which is an autonomous district. The entities and the BD BiH have their own government and assemblies and enact their laws and regulations, which are typically harmonized, but yet separate and distinct. Furthermore, the FBiH consist of 10 administrative units – cantons, which have their own constitutions and administrative organization. We can conclude that BiH has a specific and very complex constitutional and legal system which remains inefficient and is subject to different interpretations. The complicated decision-making process has contributed to delay in structural reforms and reduce the country's capacity to make progress towards the EU.25 Few credible steps have been taken to improve the adoption of key legislation relevant to EU integration. One of those steps was introduction of notary office into domestic legal system. It represents an effort of getting closer to the European standards governing legal services, especially in the fields of civil and business law.26 Before the Second World War the office of notary did exist in the region of the ex Yugoslavia but was abolished by enactment of the Implementing Statute relating to the Office of Notary passed on 17th November 1944. After abolition of the office of notary the notary's duties were at first taken over by the courts; later some of the notarial duties were also transferred to the advocates and administrative bodies.27 Although a number of notarial positions were determined by the Ordinance on the Number and Location of Notarial Positions for BiH as well, they never start with performance of the duties.28 Institution of the notary was introduced for the first time in the legal system of Bosnia and Herzegovina (BiH) in 2007. As a consequence of constitutional organization of the State, three different laws have been introduced: the Notary Law of the Federation of Bosnia and 25 European Commition, Bosnia and Herzegovina 2011 Progress Report, page 11. 26 M. Povlakić, Country Reports on Notary Service in Southeast European Countries, German Organisation for Tehnical Cooperation (GTZ) GmbH Open Regional Fund for South East Europe Legal Reform, 102. 27 E. Braniselj, Notarius International, No 3-4/2004, page 169. 28 Official Gazette of the Kingdom of Yugoslavia No 7610 413 �Herzegovina (FBHLN)29, the Notary Law of the Brčko District of Bosnia and Herzegovina (BDLN)30 and the Notary Law of the Republic of Srpska (RSLN)31. First notaries started to perform their duties in 2007 in Federation of Bosnia and Herzegovina and Brčko District, while notaries of the Republic of Srpska began their work in March 2008. The Notary as existing today in BiH belongs to the Civil law or Latin notarial system. The specific structuring of preventive justice differs from country to country. In general preventive justice include all means of reducing resort to the courts for the settlement of controversies. The term should cover the following things: - the legal settlement of issues of fact through administrative tribunals, leaving a resort to the courts on issues of law; the prevention of litigation through the settlement of disputes out of court and the prevention of disputes through care in the avoidance of grounds of disputes, when entering into transactions giving rise to legal rights.32 Notaries are part of prevetive justice and they can contribute to development of legal system in general troughout the strengthening of legal certainty, protection of public interest and avoiding disputes. 2.Entry to the profession and powers of notaries A notaries are appointed to a vacant notarial position by the Justice Minister of Federation of Bosnia and Herzegovina, Justice Minister Republic of Srpska and President of Judiciary Commission of Brčko District BiH (hereinafter: Justice Minister). Advertising of a free notarial position is announced by the Justice Ministry in the Official Gazette of the relevant entity/District. In order to be appointed as notary, a candidate must fulfill the following requirements: - Bosnian nationality, Legal capacity and sound health, An academic title as graduate in law, Successfully completed the bar exam, Successfully completed the notary exam, 29 Official Gazette of the FBiH No 42/02 30 Official Gazette of the BDBiH No 09/03 31 Official Gazette of the RS No 86/04, 02/05, 74/05, 76/05, 91/06, 37/06, 50/10 32 W. F. Dodd, Progress of Preventive Justice, American Bar Association No 6-1920, 151. 414 �- - Not to have been convicted of crimes against humanity and international law, offence against duty or any other premeditated offence that is still a matter of criminal public record with the relevant body at the time of appointment, Not to be a member of a political party.33 In accordance with the BiH Law notaries draw up authentic documents relating to legal transactions or for proceedings establishing a legal right; they take documents, money and securities for delivery to third parties or to state bodies into safekeeping and on behalf of the court or other state body handle matters which can be passed on to them in accordance with the law.34 Certain legal transactions require the form of a notarial act in order to be valid, in particular: - Contracts relating to the settlement of financial relations between spouses, Contracts relating to the disposal of the assets of a minor or persons without legal capacity, A promise of a gift, Incorporation documents for a legal entity, All types of real-estate contracts. All of the aforementioned legal transactions have to be authenticated by a notary. Transactions that are concluded without observing the statutorily or contractually required form or that are not given corresponding approval will be null and void.35 3.Functions of the notary office in BiH Notaries in BiH, like a most of notaries in civil law countries all over the world, exercise a public power. Their primary task is to confer authenticity on the legal instruments and contracts they establish for their clients, mainly in area of civil law. Although notaries in BiH are not paid by the State, this does not make their role any less of a public role. They hold a portion of public power and have the status of public official. The notary’s role contributes to preventive justice and increasing legal certainty in BiH in several ways: 33 Article 26 FLN,208 RSLN, 5 BDLN 34 Articles 69-72 FLN, 64-67 RSLN, 43-46 BDLN 35 Article 73 FLN, 68 RSLN, 47 BDLN 415 �3.1.Ensuring legal certainty Although one of the most important principle of the Contract law in BiH is freedom of contract it does not eliminate the need for supervision regarding its implementation. The notarial function is particularly important in ensuring an effective legality check.36 By placing the State's seal next to the signatures of the parties on the instruments they draw up, notaries are responsible for the content and the form. They ensure that the authentication process has been respected perfectly and that the authenticated instrument expresses the wishes of its signatories, their correct identity and the date and substance of their commitments.37Authentic instruments in general have almost same value as a judgment and can be contested only through judicial proceedings. Furthermore, each notary has a legal duty to be aware of the provisions of the 2009 Law on the Prevention of Money Laundering and Financing of Terrorist Activities (hereinafter Act) to prevent and detect the commission of money laundering and terrorist financing.38 The 2009 Act transposes the Third EU Money Laundering Directive (2005/60/EC) and associated implementing Directive 2006/70/EC into domestic legal system. When performing duties notary, if they found that there are reasons to suspect money laundering or funding of terrorist activities, in connection with transaction or certain person, they are obliged to inform the State Investigation of Protection Agency -Financial-Intelligence Department without delay (FID). Every time when a client requests an advice in relation to money laundering or funding of terrorist activities, notaries have to inform the FID immediately and not later than three working days from the date when the client requested such advice. 39 Throughout this function notaries protect not just consumers but public interest as well. 3.2.The notarial function in avoiding disputes Notaries act as independent, impartial and objective advisers to all parties to contract or a transaction. The independence and impartiality is ensured by an incompatibility of notary's 36 C. Jaquet, Notariat without borders: legal security at the service of Europeans, 1st Congress of EU Civil Law Notaries 37URL http://www.cnue.be/ 38 Official Gazette of BiH, No 53/09 39 Article 41 of the Act. 416 �work with any other for-profit work with the exception of the administration of his/her own assets. Nevertheless, a notary public may perform any scientific, publishing, teaching, interpreting, expert witnessing and artistic work against payment.40 The impartiality of the notary in all his activities is the foundation of the Notary profession in BiH. The notary must exercise his office faithfully to his oath. He is not the representative of one party, but an independent and impartial guide for the parties concerned - unlike an advocate who always looks for the benefits for his client. The impartiality of the notary guarantees a new contractual order which is characterised by the search for balance between the parties and the protection of the consumer.41 They examine the intentions of the parties, draft the contracts and instruments necessary to carry out the intended transaction and ensure that the contractual provisions are in full compliance with the law. They also verify that the parties have full capacity to enter into the intended agreement and that they have fully understood the legal implications of their commitment.42 Otherwise, the civil law notary is required by law to refuse his participation. Essential idea of notaries impartiality is to establish a preventive legal control by informing and advising clients on the legal and financial consequences of their transactions. This is why notaries are thought of as amicable settlement magistrates, practicing preventive justice.43 3.3.Realising justice Realising justice is also very important function of the notaries in BiH as well as important part of a system of preventive justice. The backlog of cases remained one of the most acute problems facing the BiH judiciary and court proceeding are generally lengthy. Despite the many reforms conducted by the State bodies, the backlog still stands at over 2.1 million cases country-wide. The fragmented legal framework across the country restricts effectiveness of 40 Article 56 FLN, 51 RSLN, 35 BDLN 41 XXIV International Congress of the Latin Notariat, Mexico City, October 2004, Impartiality of the Notary: ensuring certainty in contractual relationships, http://www.uinl.net/congreso.asp?idioma=ing&submenu=CONGRESOEJORNADAS&submenu2=CON CLUSIONESANTERIORES 42 Article 80 FLN, 70 RSLN, 53 BDLN 43 http://www.cnue.be/ 417 �judiciary system in BiH. Also existence of 14 different ministries of justice with its own budget continues to adversely affect the independence of the judiciary in BiH.44 Notarial documents enjoy a presumption of legality and exactness of content and may only be contradicted through judicial proceedings. They have evidentiary value and enforceability which reduces the costs of lawsuits as it avoids the reiteration during the proceedings of proof that had already been declared extra judicially. Enforceability avoids long and costly lawsuits and evidently represents a saving on costs. Like judicial decisions, they are enforceable, enabling the parties to have their obligations enforced directly by the judicial officers, without having to pass before the courts. 4.CONCLUSION By introducing the notary profession, BiH has made a big step towards improvement of consumer protection, independent and impartial judicial system, as well as adaptation domestic legal system with European Union law. In our study we have found that notaries have contributed to the development of preventive justice in BiH by ensuring legal certainty, avoiding disputes and realising justice. As independent, impartial and objective advisers to all parties to a transaction, notaries lead to the strengthening of legal certainty and protection of public interest. They also provide the market and development with trust. The main idea by introducing notary office into the legal system of BiH was to establish a preventive legal control in order to avoid costly and time-consuming litigation. Additional improvement in the context of the administration of preventive justice can be made by transferring more powers to the notary office, such as process of mediation which makes process of dispute resolution simpler, quicker and less costly in the interest of citizens. REFERENCES C. Jaquet, Notariat without borders: legal security at the service of Europeans, 1st Congress of EU Civil Law Notaries E. Braniselj, Notarius International, No 3-4/2004 European Commition, Bosnia and Herzegovina 2011 Progress Report High Judicial and Prosecutorial Council of Bosnia and Herzegovina, 2010 Annual Report 44 High Judicial and Prosecutorial Council of Bosnia and Herzegovina, 2010 Annual Report, 144. 418 �Law on the Prevention of Money Laundering and Financing of Terrorist Activities, Official Gazette of BiH, No 53/09 M. Povlakić, Country Reports on Notary Service in Southeast European Countries, German Organization for Technical Cooperation (GTZ) GmbH Open Regional Fund for South East Europe - Legal Reform Notary Law of the Brčko District of Bosnia and Herzegovina, Official Gazette No 09/03 Notary Law of the Federation of Bosnia and Herzegovina, Official Gazette No 42/02 Notary Law of the Republic of Srpska, Official Gazette No 86/04, 02/05, 74/05, 76/05, 91/06, 37/06, 50/10 Ordinance on the Number and Location of Notarial Positions, Official Gazette of the Kingdom of Yugoslavia No 7610 W. F. Dodd, Progress of Preventive Justice, American Bar Association No 6-1920 XXIV International Congress of the Latin Notariat, Mexico City, October 2004, Impartiality of the Notary: ensuring certainty in contractual relationships, http://www.uinl.net/congreso.asp?idioma=ing&submenu=CONGRESOEJORNADAS&subm enu2=CONCLUSIONESANTERIORES http://www.cnue.be/ 419 � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Extent The size or duration of the resource. 1197 Title A name given to the resource Civil Law Notaries in Bosnia and Herzegovina: Actors in Preventive Justice Author Author Bakšić, Šukrija Oruč, Esad Abstract A summary of the resource. Civil law notaries are professional lawyers and public officials appointed by the State to confer authenticity on legal deeds and contracts contained in documents drafted by them and to advise persons who call upon their services. Institution of the notary was introduced for the first time in the legal system of Bosnia and Herzegovina in 2007. Introduction of the office of notary was one of the steps taken to ensure independent and impartial judiciary and to adapt legal system with European Union law. Before its introducing there was no institution or legal profession which acted impartially on behalf of all parties to a contract or transaction. Notarial services are very wide and complex. It encompasses all judicial activities in noncontentious matters, ensure legal certainty to clients, thus averting disputes and litigation. As a guarantor of legal certainty, notary is one of the most important actors of preventive justice which include all means of reducing resort to the courts for the settlement of controversies. In this study we analyzed contribution of notary office to preventive justice in Bosnia and Herzegovina. Keywords: civil law notary, preventive justice, legal certainty, realising justice, avoiding disputes Date A point or period of time associated with an event in the lifecycle of the resource 2012-05-31 Keywords Keywords. Conference or Workshop Item PeerReviewed H Social Sciences (General),T Technology (General) https://omeka.ibu.edu.ba/files/original/420badc84095e83d23957322ebbf4162.pdf f0ddc75bfa35c7a190eeb4c650e4eccf PDF Text Text Zhu, Jiang, FeiXiong, DongzhenPiao, Yun Liu, Ying Zhang (2011). “Statistically Modeling the Effectiveness of Disaster Information in Social Media”, 2011 IEEE Global Humanitarian Technology Conference, s. 431-436. Does predefined erp implementation methodology work for public companies in transitioning country? Classification of EEG signals for epileptic seizure prediction using ANN JasminKevric, AbdulhamitSubasi International Burch University, Faculty of Engineering and Information Technologies, FrancuskeRevolucije bb, Ilidža, Sarajevo, 71210, Bosnia and Herzegovina. E-mails:jkevric@ibu.edu.ba, asubasi@ibu.edu.ba Abstract In this paper, we developed a model for classification of EEG signals. The aim of the study is to determine whether this model can be used for epileptic seizure prediction if “pre-ictal” stages were successfully detected. We analyzed long-term Freiburg EEG data. Each of 21 patients contains datasets called “ictal” (seizure) and “inter-ictal” (seizure-free). We extracted 4096-samples (or 16 seconds) long segments from both datasets of each patient. These segments were decomposed into time-frequency representations using Discrete Wavelet Transform (DWT). The statistical features from the DWT sub-bands of EEG segments were calculated and fed as inputs to Multilayer Perceptron (MLP) and Radial Basis Function (RBF) network classifiers using 10-fold cross validation. We also applied multiscale PCA (MSPCA) de-noising method to determine if it can further enhance the classifiers’ performance. MLP-based approach outperformed RBF classifier with or without MSPCA, which significantly improved the classification accuracy of both classifiers. The proposed MLP-approach with MSPCAachieved a classification accuracy of 95.09%. We showed that a high classification accuracy of EEG signals can be accomplished in cases when additional “pre-ictal” class is introduced. Therefore, the proposed approach may become an efficient tool to predict epileptic seizures from EEG recordings. Keywords: Electroencephalogram (EEG); Epileptic seizure; Discrete Wavelet Transform (DWT); Multilayer Perceptron (MLP); Radial Basis Function (RBF) network; Multiscale PCA (MSPCA); Machine learning. 491 �1.INTRODUCTION Noninvasive electrodes on the scalp can record the brain's electrical activity called as electroencephalogram (EEG), produced by billions of neurons firing within the nervous system. The EEG signal is characterized by a nonstationarity in the waveforms and semistationary time-dependent states, and detection of these characteristics is a difficult task (Bigan, 1998). Over 50 million people in the world are affected by the epilepsy, the second mostcommon neurological disorder after stroke (D’Alessandro et al., 2003). Abnormal movements and seizures, resulting from the brain cells' excessive electrical discharge, are the signs of epilepsy. One of the most important causes of stress, morbidness and anxiety in epileptic patients is the inability of predicting seizure onset (Murray, 1993; Buck et al., 1997). Thereliable predictability of seizure onset would dramatically improve the safety and quality of life of these patients who cannot be treated successfully by common therapeutic options (Schachter, 1994). For example, patients would be able to prevent dangerous situations when being warned of upcoming seizure. Various automated intervention systems and measures could be implemented like applying electrical brain stimulations or delivering short-acting anticonvulsant drugs by using implanted devices (Stein et al., 2000;Elger, 2001). Additionally, the investigation of the pathophysiological mechanisms causing seizures could be improved by the accurate detection of states preceding seizures. Mormann et al., (2007) stated that seizure prediction is the long and winding road in their review article. D’Alessandro et al., (2003)used intelligent genetic search technique to classify preseizure and non-preseizure classes from four patients by a probabilistic neural network, reporting a sensitivity of 62.5% with 90.5% specificity. Costa et al., (2008)compared 6 types of neural network architectures which used 14 features extracted from EEG of two patients to classify brain states into four classes: inter-ictal, pre-ictal, ictal and pos-ictal. The accuracies of up to 99% were achieved. Mirowski et al., (2009)achieved 71% sensitivity and 0 false positives using convolutional networks combined with wavelet coherence. Chisci et al., (2010) used Autoregressive (AR) models to classify pre-ictal and inter-ictal classes from nine patients, reporting 100% sensitivities and average false positive rates of 0.174/h (on the interictal dataset). This paper is organized as follows. Section 2 describes the EEG data, signal processing and feature extraction methods, and the artificial neural networks with a brief description of each one. In section 3, the performance of the proposed system is presented and discussed. Finally, section 4 presents concluding remarks and perspectives for future work. 492 �2.Materials and methods 2.1 Subjects and data recording We analyzed long-term EEG data recorded during invasive pre-surgical epilepsy monitoring at the Epilepsy Center of the University Hospital of Freiburg, Germany. The Neurofile NT digital video EEG system with 128 channels, 256 Hz sampling rate, and a 16 bit analogue-todigital converter was used to acquire the EEG data. Each of 21 patients, suffering from medically intractable focal epilepsy, contains datasets called “ictal” and “inter-ictal”. The “ictal“ dataset consists of files containing epileptic seizures, each having a seizure-free "preictal" period of at least 50 minutes. The “inter-ictal“ dataset consist of approximately one day of seizure-free EEG recordings for each patient. Each patient had between two and five seizures, with an average of 4.2 seizures per patient or a total number of 87 seizures(Maiwald et al., 2004).The onset and end times of each seizure were determined by visual examination of skilled epileptologists. 2.2 Multiscale Principal Component Analysis Multiscale Principal Component Analysis (MSPCA) combines the wavelet analysis with PCA. The MSPCA method incorporates the decomposition of each variable on a selected family of wavelets during which the wavelet coefficients are thresholded. After that, the PCA model is separately built for the coefficients at each scale. In order to yield one model for all scales together, the models at important scales, which show process disturbances or abnormal operation, are merged in an effective scale-recursive way(Bakshi 1998; Ganesan, Das, & Venkataraman, 2004).Because of its multiscale type, it is suitable to use MSPCA for modeling of data consisting of contributions from events which behavior changes over time and frequency. MSPCA is powerful tool for monitoring autocorrelated measurements without time-series modeling or matrix augmentation due to approximate decorrelation of wavelet coefficients. The MSPCA method not only selects and monitors the significant signal features but also conforms to the nature of the signal (Bakshi 1998). 2.3 Discrete Wavelet Transform Signals like EEG may contain transitory or non-stationary characteristics. That is why Fourier Transform, which can be applied to the stationary signals, is not an ideal method to be directly applied to signals like EEG. Therefore, time-frequency methods like Wavelet Transform should be used. The analysis based on Discrete Wavelet Transform is best explained in terms of filter banks. Multi-resolution decomposition of a signal is the procedure of using a group of filters to separate that signal into various spectral components. Every stage of this procedure consists of two digital filters and two down-samplers by 2. The first filter is the discrete mother wavelet, being high-pass in nature. The second filter is its mirror version, being low pass in 493 �nature. Outputs of the first high-pass and low-pass filters, once being down-sampled, provide the detail D1 and the approximation A1, respectively (Adeli, Zhou, & Dadmehr, 2003;Marchant, 2003; Semmlow, 2004). In DWT analysis it is very important to choose the appropriate number of decomposition levels and appropriate wavelet selection. The components of the dominant frequency of the signal are the main base for choosing the number of decomposition levels. Distribution of energy of the EEG signal in frequency and time is shown by a compact representation of the extracted wavelet coefficients. Using statistics over the wavelet coefficients sets helped in decreasing the dimensionality of the extracted feature vectors (Kandaswamy et al., 2004).Subasi (2007) and Subasi&Gursoy (2010) achieved high accuracies in classifying EEG signals using statistical feature vectors extracted from wavelet coefficients. 2.4 Multilayer Perceptron Multilayer feedforward networks is composed of a set of source nodes which serve as sensory units that form the input layer, one or more hidden layers and an output layer. Hidden layers and an output layer consist of computational nodes. The input signal is transmitted through the network in a forward direction, layer by layer. This type of neural networks, which represents a generalization of the single-layer perceptron, is generally known as multilayer perceptron (MLP). When trained in a supervised manner using highly popular and computationally efficienterror back-propagation algorithm, multilayer perceptrons can successfully solve complex and different problems, but certainly do not provide an optimal solution for all solvable problems. Essentially, error back-propagation learning consists of a forward pass and a backward pass. In the forward pass, the effect of an input vector, when being applied to the sensory nodes, propagates through the network. At the end, a set of outputs, as the real response of the network, is formed. The synaptic weights are all fixed during this stage. However, these synaptic weights are being tuned according to errorcorrection rule during the backward pass. Namely, an error signal is produced as the real response of the network is subtracted from a desired (target) response. This error signal is then propagated backward through the network during which the synaptic weights are tuned so that the difference between the real and the desired response of the network decreases.One or more layers of hidden neurons enhance network’s learning of difficult problems by extracting more significant features from the input vectors(Haykin 1999). 2.5 Radial Basis Function Network The design of a neural network can also be perceived as a curve-fitting (approximation) problem in a high-dimensional space,where learning is viewed as finding a surface which 494 �represents a best fit to the training data in a multidimensional space. This multidimensional surface is then used to interpolate the test data. The method of radial-basis functions is motivated by such a viewpoint. The early work on radial-basis functions is reviewed in Powell (1985).A radial-basis function (RBF) network basically consists of three layers having completely different tasks. The input layer connects the network to the environment via source nodes that serve as sensory units. A nonlinear transformation from the input space to the hidden space of high dimensionality is applied in the second layer as the only hidden layer in the network. The output layer, producing the response of the network to the input vector, is linear. The effect of applying nonlinear transformation prior to a linear transformation is explained by Cover (1965). As stated by him, there is a higher change of a pattern recognition problem to be linearly separable in a high-dimensional space. Therefore, the dimension of the hidden space in an RBF network is often made high. Moreover, the higher the dimension of the hidden space, the more accurate the approximation of smooth mapping is(Mhaskar, 1996; Niyogi and Girosi, 1996). 3. Experimental results and discussion 3.1 Experiment Classification of EEG signals consists of data acquisition and preparation, signal processing, feature extraction and classification. We propose a method based on MSPCA for denoising, DWT for feature extraction and ANNs for classification.We extracted 4096-samples-long segments from both datasets of each patient. Approximately two segments per hour were extracted from “inter-ictal” dataset, producing 1050 inter-ictal segments. We also extracted two types of segments from “ictal” dataset: ictal and pre-ictal. We used minimum number of 4096-samples-long segments to cover all 87 seizure activities, producing 652 ictal segments. We extracted five segments within a seizure-free "pre-ictal" period of 50-60 minutes, producing 435 pre-ictal segments. Only one out of six channels was used for extraction of EEG segments, although results from the different authors presented a poor performance of univariate measures (Mormann et al., 2005). We selected the number of decomposition levels for DWT to be 5 since EEG signals contain no useful frequency components above 30 Hz, and because of 256Hz sampling rate of Neurofile NT used to acquire the EEG data. Daubechies 4 (DB4) wavelet filter was used to reconstruct the detail and approximation records.All 2137 EEG segments, which belong to three different classes, were divided into sub-band frequencies A5 (0-4 Hz), D5 (4-8 Hz), D4 (8-16 Hz), D3 (16-32 Hz), D2 (32-64 Hz) and D1 (64-128 Hz). Sub-band frequencies A5 and D3-D5 almost perfectly correspond to δ (0-4 Hz), θ (4-8 Hz), α (8-12 Hz) and β (12-26 Hz) frequencies of EEG signals (Bylsma et al., 1994). A set of fifteen statistical features was then extracted from the wavelet coefficients representing these sub-band frequencies and fed as inputs to classifiers. A Multiscale PCA (MSPCA) de-noising method was also applied to determine if it can further enhance the 495 �classifiers’ performance. We implemented a classification system based on MLP and RBF network using wavelet statistical features as inputs and 10-fold cross validation method, to guarantee validity of the results. 3.2 Results We performed two types of experiment: with and without MSPCA de-noising method applied.In Table 1, we have seen that MSPCA drastically improved the classification accuracy of both classifiers, while MLP network achieved higher total classification accuracy than RBF network. The accuracies for each class are also presented in Table 1. Accuracy Accuracy Accuracy Total (Pre-ictal) (Inter-ictal) (Ictal) Accuracy MLP +DWT 2.76 % 89.43 % 60.58 % 62.99 % RBFN +DWT 7.13 % 90.57 % 54.45 % 62.56 % MLP + MSPCA+DWT 87.82 % 97.43 % 96.17 % 95.09 % RBFN + MSPCA+DWT 71.49 % 97.14 % 94.02 % 90.97 % Classifier Table 1. Accuracies of MLP and RBF network classifiers with and without MSPCA. MSPCA significantly improved the classification accuracy for ictal and pre-ictal samples, while accuracy performance for inter-ictal class was only slightly improved. Classifiers are totally useless for seizure prediction if MSPCA is not applied. 3.3 Discussion The experiment results show that MSPCA is an effective denoising method for improving the classification performance. Without MSPCA, our method classified many pre-ictal/ictal data samples as being inter-ictal.Aminghafari, Cheze, & Poggi, (2006) showed that de-noised signals by MSPCA magnify the spikes more clearly. Therefore, MSPCA enhanced our classifier's performance for about 50%. Our approach outperformed the one explained in D’Alessandro et al., (2003). Theyalso used data of only four patients to developfour different classifiers for each patient. Although Costa et al., (2008)introduced one more class (pos-ictal) and achieved accuracies of 99%, using data of only two patients from Freiburg database is insufficient to successfully train and develop a model. Mirowski et al., (2009) predicted all seizures without false positives for 15 patients, without mentioning how classifier performed on data belonging to six remaining patients 496 �from Freiburg database. Thus, sensitivity of 71% is reported, which is lower than classification accuracies for pre-ictal class of both of our classifiers. Chisci et al., (2010)used nine patients for which additional electro-corticographic recordings (grid–strip electrodes) were available and achieved 100% sensitivity with low false positive rates. However, they developed patient-specific system by training nine classifiers, where each classifier used train and test data of only one patient. Our proposed system is more general because only one classifier is developed for all patients and it is not bound to specific group of epileptic patients. 4. CONCLUSION We showed that a high classification accuracy of EEG signals can be accomplished in cases when additional “pre-ictal” class is introduced. Many research papers showed that DWT coefficients well represent the EEG signals and ensure a good differentiation between classes. However, we managed to achieve high accuracies only when MSPCA de-noising method was applied to Freiburg dataset. The accuracy may be further improved by applying dimension reduction or feature selection methods like ICA or LDA on the feature vectors. Measures that characterize the relations between two or more channels can be used to further enhance the performance. Using only inter-ictal and pre-ictal samples to train the classifier could be investigated since our aim is not seizure detection. Freiburg dataset can serve as a challenge for trying other feature extraction methods rather than DWT. The proposed approach may become an efficient tool to predict epileptic seizures from EEG recordings. REFERENCES Adeli, H., Zhou, Z., & Dadmehr, N. (2003). Analysis of EEG records in an epileptic patient using wavelet transform. Journal of Neuroscience Methods 123, 69-87. Aminghafari, M., Cheze, N., & Poggi, J.-M. (2006). Multivariate denoising using wavelets and principal component analysis. Computational Statistics & Data Analysis, 50, 2381-2398. Bakshi, B. R. (1998). Multiscale PCA with Application to Multivariate Statistical Process Monitoring. AlChE, 44(7), 1596-1610. Bigan, C. (1998). A recursive time-frequency processing method for neural networks recognition of EEG seizures. In E. C. Ifeachor, A. Sperduti, & A. Starita (Eds.), Neural Networks and Expert Systems in Medicine and Healthcare. Singapore: World Scientific. Buck, D., Baker, G. A., Jacoby, A., Smith, D. F., & Chadwick, D. W. (1997). Patients' experiences of injury as a result of epilepsy. Epilepsia, 38, 439-44. Bylsma, F., Peyser, C., Folstein, S., Ross, C., & Brandt, J. (1994). EEG power spectra in Huntington’s disease: clinical and neuropsychological correlates. Neuropsychologia 32(2), 137-150. 497 �Chisci, L., Mavino, A., Perferi, G., Sciandrone, M., Anile, C., Colicchio, G., et al. (2010). Real-Time Epileptic Seizure Prediction Using AR Models and Support Vector Machines. IEEE Transactions on Biomedical Engineering, 57(5). Costa, R. P., Oliveira, P., Rodrigues, G., Leitao, B., & Dourado, A. (2008). Epileptic Seizure Classification Using Neural Networks with 14 Features. In I. Lovrek, R. J. Howlett, & L. C. Jain (Eds.), KES 2008, Part II, LNAI 5178 (pp. 281-288). Springer-Verlag Berlin Heidelberg. Cover, T. M. (1965). Geometrical and Statistical properties of systems of linear inequalities with applications in pattern recognition. IEEE Transactions on Electronic Computers, 326334. D'Alessandro, M., Esteller, R., Vachtsevanos, G., Hinson, A., Echauz, J., & Litt, B. (2003). Epileptic seizure prediction using hybrid feature selection over multiple intracranial EEG electrode contacts: a report of four patients. IEEE Transactions on Biomedical Engineering 50 (5), 603-615. Elger, C. E. (2001). Future trends in epileptology. Current Opinion in Neurology, 14, 185186. Ganesan, R., Das, T. K., & Venkataraman, V. (2004). Wavelet-based multiscale statistical process monitoring: A literature review. IIE Transactions, 36(9), 787-806. Haykin, S. (1999). Neural Networks: A Comprehensive Foundation (Second ed.). Prentice Hall. Kandaswamy, A., Kumar, C., Ramanathan, R., Jayaraman, S., & Malmurugan, N. (2004). Neural classification of lung sounds using wavelet coefficients. Computers in Biology and Medicine, 34(6), 523-537. Maiwald, T., Winterhalder, M., Aschenbrenner-Scheibe, R., Voss, H. U., Schulze-Bonhage, A., & Timmer, J. (2004). Comparison of three nonlinear seizure prediction methods by means of the seizure prediction characteristic. Physica D 194, 357-368. Marchant, B. P. (2003). Time–frequency analysis for biosystem engineering. Biosystems Engineering, 85(3), 261-281. Mhaskar, H. N. (1996). Neural networks for optimal approximation of smooth and analytic functions. Neural Computation, 8, 164-177. Mirowski, P., Madhavan, D., LeCun, Y., & Kuzniecky, R. (2009). Classification of patterns of EEG synchronization for seizure prediction. Clinical Neurophysiology, 120(11), 19271940. Mormann, F., Andrzejak, R. G., Elger, C. E., & Lehnertz, K. (2007). Seizure prediction: the long and winding road. Brain, 130, 314-333. Murray, J. (1993). Coping with the uncertainty of uncontrolled epilepsy. 2, 167-178. 498 �Niyogi, P., & Girosi, F. (1996). On the relationship between generalization error, hypothesis complexity, and sample complexity for Radial Basis Functions. Neural Computation, 8, 819842. Powell, M. J. (1985). Radial basis functions for multivariable interpolation: a review. IMA Conference on Algorithms for the Approximation of Functions and Data. Shrivenham: RMCS. Schachter, S. C. (1994). The brainstorms companion: epilepsy in our view. New York: Raven Press. Semmlow, J. L. (2004). Biosignal and biomedical image processing: MATLAB-based applications. New York: Marcel Dekker, Inc. Stein, A. G., Eder, H. G., Blum, D. E., Drachev, A., & Fischer, R. S. (2000). An automated drug delivery system for focal epilepsy. Epilepsy Res, 39, 103-114. Subasi, A. (2007). EEG signal classification using wavelet feature extraction and a mixture of expert model. Expert Systems with Applications 32, 1084-1093. Subasi, A., & Gürsoy, M. I. (2010). Comparison of PCA, ICA and LDA in EEG signal classification using DWT and SVM. Expert Systems with Applications 37, 8659-8666. Classification of Fetal State from the Cardiotocogram Recordings using ANN and Simple Logistic Hakan Sahin, Abdulhamit Subasi International Burch University, Faculty of Engineering and Information Technologies, 71000, Sarajevo, Bosnia and Herzegovina E-mail:hakanshah@hotmail.com , asubasi@ibu.edu.ba Abstract In this study, we present a comparison of machine learning technics using antepartum cardiotocographs performed by SisPorto 2.0 in predicting newborn outcome. CTG is widely used in pregnancy as a technique of measuring fetal well-being, mainly in pregnancies with increased risk of complications. It is a non-invasive way for checking the fetal conditions in the antepartum period. CTG is a continuous electronic record of the baby’s heart rate acquired via an ultrasound transducer placed on the mother’s abdomen. The information 499 � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Extent The size or duration of the resource. 1208 Title A name given to the resource Classification of EEG signals for epileptic seizure prediction using ANN Author Author Kevric, Jasmin Subasi, Abdulhamit Abstract A summary of the resource. In this paper, we developed a model for classification of EEG signals. The aim of the study is to determine whether this model can be used for epileptic seizure prediction if “pre-ictal” stages were successfully detected. We analyzed long-term Freiburg EEG data. Each of 21 patients contains datasets called “ictal” (seizure) and “inter-ictal” (seizure-free). We extracted 4096-samples (or 16 seconds) long segments from both datasets of each patient. These segments were decomposed into time-frequency representations using Discrete Wavelet Transform (DWT). The statistical features from the DWT sub-bands of EEG segments were calculated and fed as inputs to Multilayer Perceptron (MLP) and Radial Basis Function (RBF) network classifiers using 10-fold cross validation. We also applied multiscale PCA (MSPCA) de-noising method to determine if it can further enhance the classifiers’ performance. MLP-based approach outperformed RBF classifier with or without MSPCA, which significantly improved the classification accuracy of both classifiers. The proposed MLP-approach with MSPCAachieved a classification accuracy of 95.09%. We showed that a high classification accuracy of EEG signals can be accomplished in cases when additional “pre-ictal” class is introduced. Therefore, the proposed approach may become an efficient tool to predict epileptic seizures from EEG recordings. Keywords: Electroencephalogram (EEG); Epileptic seizure; Discrete Wavelet Transform (DWT); Multilayer Perceptron (MLP); Radial Basis Function (RBF) network; Multiscale PCA (MSPCA); Machine learning. Date A point or period of time associated with an event in the lifecycle of the resource 2012-05-31 Keywords Keywords. Conference or Workshop Item PeerReviewed H Social Sciences (General),T Technology (General) https://omeka.ibu.edu.ba/files/original/eb4e2e28f478439770947f4f88e915f9.pdf 606ae5e9ebc595004cc58b8adad35fe9 PDF Text Text Classification Of Emg Signals Using Decision Tree Methods Selami Keleş, Abdulhamit Subaşı E-mail:keles_selami@yahoo.com, asubasi@ibu.edu.ba Abstract Nowadays, Usage of EMG signals are increasing very fast among the Medical Professionals to determine specific disorders. Recent Computational Intelligence studies show that EMG signals can be processed by machine learning methods. The aim of our study is to implement an accurate system to classify EMG signals using decision tree algorithms. We preprocessed the EMG signals and used autoregressive method (AR) for feature selection. Features are reduced by different filtering methods and applied to decision tree classification algorithms, namely Simple CART, C4.5, Random Forest and Random Tree. EMG signals are classified as Myopathy, Neuropathy and Normal. All the data are compared each other on the table try to find out the best classification and feature reduction methods. While tree algorithms classify the data with the accuracy between %89, 82 and %99, 25, feature reduction slightly affects the accuracy of the classification methods. It has been shown that a successful automatic diagnostic system implemented to classify EMG signals by using decision tree algorithms. Furthermore, future reduction may help to increase the accuracy of the system. Keywords: EMG, Neuropathy, Myopathy, Simple CART, C4.5, Random Tree, Random Forest, Feature reduction. 1.INTRODUCTION Early and accurate diagnosis is important for neuromuscular diseases that help the patient to get full recovery or have better health after therapy. Sometimes, clinical examination is not enough to diagnose and to find the location of disorders [1]. Therefore, it has high importance to find correct location of the disorders to accurate diagnosis and therapy. EMG recordings are more useful than clinical examination to find out the muscle fibers involved in a disorders and abnormal sensory nerve conduction. It allows the clinician to diagnosis without needing a muscle biopsy and raises the clinician response time and helps to treat some disorders. The analysis of EMG signals can be done only by qualified and professional neurologist. The problem is that, there are few professionals to interpret the EMG waveforms and use the necessary techniques. Therefore, it is important to develop an automated diagnostic system by using EMG signals. The application of Computational Intelligence (CI) techniques can be 354 �used to develop an automated diagnostic system that detects and classify the neuromuscular diseases by processing EMG signals which helps the neurologists to diagnose the neuromuscular anomalies. The MUP assessment may not be satisfactory to detect small deviation or miscellaneous patterns of abnormalities [1]. Therefore, to design an accurate automatic EMG signal classification system, different EMG analysis algorithms have been developed[2, 3] To develop an intelligent diagnostic system, fist, EMG signals have to be pre-processed and extracted the characteristic information. Then, extracted features that contain the time and frequency domain information, processes by using wavelet coefficients, Fourier coefficients, autoregressive coefficients or other signal processing techniques. After all, processed information can be used as input to the classifier such as NNs, SVM or Decision Tree to classify the disease. One of the most popular MachineLearning Method ANN has been widely used to classify the EMG data. In order to increase the classification success, ANN can combine the best of both time and frequency domain measures, but it is not enough for clinical use [4, 5]. Christodoulou and Pattichis used Self Organized Feature Maps and Learning Vector Quantization used to classify MUP’s [5]. Genetic algorithms were used by Schizas ve Pattichis to classify the EMG signals [6]. Multilayer Perceptron Neural Networks (MLPNN), Dynamic Fuzzy Neural Network (DFNN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) based classifiers were compared by Subaşı. ANFIS model has reported more successful than others with the accuracy of 95%. [7]. SVM classifier is used by Katsis at. al. and the classify the EMG signals whit the correct identification rates of 93, 95 and 92% for normal, myopathy and neuropathy, respectively [8]. The result of another comparisonresearch between Combined Neural Network (CNN) and Feedforward Error BackpropagationANN (FEBANN) classifiers was describedby Bozkurt. Even the CNN didn’t provide the fast enough classification;itgaveslightly higher success than the FEBANN with the accuracy of 92% [9] There are still challenges to develop an accurate and practical automated system. EMG signals vary patient to patient in a very large range. Signal amplitude and duration changes by patient age. This problem can be solved by designing a signal processing techniques that conserve or capture distinctive information in raw EMG readings. High-quality set of features[10]. 2.EMG EMG can be defined as a method of analyzing neuromuscular conditionsdepends on cell action potentials for the duration of muscle action. The specification of the EMG signal is 0.01-10mV and 10-2000Hz on average. This signal has information about location, reason of disorder and type of illness. For example, while the EMG pulse duration shows the location and metabolic condition of the muscle [11],odd spikes may point to the myopathy. 355 �Electromyograph records the Motor Unit Action Potential (MUAP). EMG can be categorized into needle or fine wire EMG and surface (sEMG). While EMG signals are recording, some instruments are required including, electrodes, a signal acquisition system and signal filters. Generally, EMG instruments are produced with typical settings for signal characteristics such as filter bandwidth, gain and input impedance [12]. The needle electrode or wire electrode can reach the individual motor unit and get the action potential more accurately than the surface EMG.Surface EMG electrode is more useful than needle or wire electrodes, because it is used by attaching the body instead of inserting anything in it. EMG signals are recorded at hospital lab by Electromyographers[10]. 3. Myopathy Myopathy is a muscle disorder especially skeletal muscle, which is caused by several reasons such as injury of muscle group or some genetic mutation. It obstructs the proper tasks of muscle fibers. The patient suffering with myopathy has weak muscle and has difficulties to perform regular tasks. Depending on the severity of disease, sometimes it is impossible to make any movement by using affected muscle. There are a number of types of myopathy including; Muscular dystrophy, Congenital muscular dystrophy, Duchenne muscular dystrophy, Becker muscular dystrophy, Emery–Dreifuss muscular dystrophy, Myotonic muscular dystrophy, Distal muscular dystrophy, limb–girdle muscular dystrophy, facioscapulohumeral muscular dystrophy and oculopharyngeal muscular dystrophy [10] Neuropathy Simply, Neuropathy is the term for describing damage to nerves of nerves system. It causes pain and some disability. Neuropathy can be caused by variety of precipitating factors including infection, diabetes; alcohol abuse, cancer chemotherapy and injury. When a single nerve is affected, it is called Mono-neuropathy. When a group of nerves or all nerves of peripheral nerve are affected, it is called Polyneuropathy. Poly neuropathies are similar because of inadequate manner in which sensory nerves react to malfunction. EMG diagnosis is not considerably useful for Polyneuropathy, because the patients with polyneuropathy have normal electrophysiological characteristics [10] Decision Tree Classifiers; The Decision Tree is a classification algorithm thatclassifies a pattern by asking questions, in which the next question asked depends on the answer to the present question [13].It uses a “divide-andconquer” approach to solve the learning problems [14]. Decision Tree learning methods are one of the most popular inductive inference algorithms and 356 Figure-1 �have been used a wide range of task about medical diagnosis [15]. The instances are classified by sorting them down the tree from root to some leaf node which the classification is provided in decision tree algorithms. The attributes of the instance are tested at each node and sent to the sub node or leaf node from one of the branch which correspond the possible values of that attribute [15]. The numeric attributes are tested by comparing a pre-defined constant value at the node and it gives two or three-way split depends on the several different possibilities. [14]. Trained trees can be shown by a set of ifthan rules to increase human readability [15]. An example of three is shown in the figure-1 which is adopted from Quinlan research [16]. 4.C4.5 C4.5 is develop by Ross Quinlan [17] to make complex decision trees more understandable by using a list of rules of the form “If X and Y and Z and ….then class A” where rules are grouped together for each class. When the first rule is found which satisfies the condition of case, the instance is classified. If there is no rule which is satisfied by the case, it is sent to default class. The basic disadvantage of the C4.5 algorithm is requirement of high amount of CPU time and system memory[18]. 5.Random Tree Random Decision Tree is a randomly trained ensemble of decision trees which is proposed by Fan et al. [19]. The features are randomly selected at each node, while training trees phase is proceeding. A selected discrete feature never selected again till it is vain to use the same discrete feature more than once.Conversely, it is possible to choose continues features several times as long as every time, using randomly selected splitting value. Each tree gives raw posterior probabilities at the classification phase and outputs of each tree in the ensemble are averaged for last posterior profanities estimation. It is proofed that the Random Decision Tree is highly accurate classification method for both 0-1 loss and cost-sensitive loss function. [20]. 6.Random Forest Random forest is a tree algorithm which composed of a number of tree predictors. In this algorithm, each tree is shown by a random vector which is independently taken from the same distribution in the forest. As the number of the tree increase in the forest, the generalization error converges to a limit. The strength of the individual tree and relationship between the trees affects the generalization error. Once all trees in the forest produce a result, they are voted for the most passible class [21]. It is one of the most successfulclassification 357 �methods among the available algorithms for many data type [22], but opposite to other decision tree methods, it makes classification which is difficult to deduce by human [23]. 7.Classification and Regression Trees(CART) Classification and Regression Trees (CART), proposed by Breiman at al., [24], was a revolutionary improvement of Machine Learning and Data Mining fields which can be used almost any domain such as electrical engineering, biology and medical researches. It is a binary repeated division process which can work with the nominal and continues data. The raw form of data is processed without requiring binning. The growing trees aren’t halted by using stooping rules till it reaches maximum size and then clipped back to the root by cost-complexity pruning method. The pruned next split contributes the overall performance of the tree. The CART algorithm is projected to grow a sequence of nested pruned trees that all of them are nomine of the optimal trees. To find out the “right sized” tree, the predictive success of every tree is evaluated at the pruning process. The performance of the tree is measured by test data or cross validation method and tree is selected after evolution, because CART doesn’t have any internal performance measurement method depending on training data. [18] 8.AR model An Autoregressive (AR) model is used to estimate the different kinds of naturel fact in signal processing and statistic fields which were originally proposed by Yule. It contains a set of linear estimation formulas which is used to predict the output of a system depends on the previous output. [25, 26] There are a number of methods to estimate the AR model parameters. Some of them are the Yule-Walker, Burg(1968), covariance and modified covariance methods. It is easy to access and use these methods in many software packages such as MATLAB (http://www.mathworks.com/products/matlab/) and Signal Processing Toolbox. The Yule-Walker technique is based on a partial form of the autocorrelation approximate to guarantee a positive semi defined autocorrelation matrix. Alternatively, the Burg method uses a form of order-recursive least square method which approximates the parameters by minimizing errors of the linear system. [10] 9.Feature selection algorithms An important issue is handling irrelevant features in pattern recognition field. Feature Selection (FS) method is necessary to find out the important features to classify the data accurately, because it was not considered how to overcome a large amount of irrelevant feature in many pattern recognition methods, while they were designing. [27,28,29] Mostly, the feature selection methods are used to increase the model performance, to abstain the 358 �overfitting, to get faster and more cost effective models and to understand the processes which produce the data. Beside the advantages, FS methods add new complexity layer to the models. [30], searching the optimal subset of relevant features. FS methods can be grouped in to three categories by the way of allying relevant features search with building classification model; filter methods, wrapper methods and embedded methods. [31] 10.Materials and Methods 10.1.Subjects and Data Acquisition The patients which samples are taken from and the control group were chosen at Neurology Department of University of Gaziantep. Measurements are taken by an EMG system (Keypoint; Medtronic Functional Diagnostics, Skovlunde, Denmark) with standard settings. The signal was obtained from biceps brachii muscle by using a concentric needle electrode (0.45 mm diameter with a recording surface area 0.07 mm2; impedance at 20 Hz below 200 KΩ). 5 Hz to 10 KHz band-pass filter was applied to the raw signal and sampled at 20 KHz for 5 s with 12-bit resolution. Then 8 KHz low-passed filter was applied. The signals are recorded from three to five different points in muscle for standardization. And also the needles are inserted in to muscle until it reaches the medial or posterior border of the muscle (at least 3-5 mm deep). The needles are moved 3-5 mm to ensure to record different MUPs at every recoding season. The signals were taken from the biceps brachii muscle of the patients under isometric condition at just about 30% of Maximum Voluntary Contraction (MVC). Before the patient diagnosis, general examination and clinical history of the patient were considered and EMG and nerve conduction tests were regarded. Unless, the EMG diagnosis results were uncertain and some other clinical reason; the muscle biopsies were not done. The data which was used for this study were collected from 27 different subjects and analyzed. Details about the subject are given below as in [3]    7 healthy subjects, (3 males, 4 females,) ages between 10 to 43 years (mean age±standard deviation (S.D.): 30.2±10.8 years) 7 myopathic subjects (4 males, 3 females) ages between 7 to 46 years, (mean age±standard deviation (S.D.): 21.5±13.3 years) 13 neuropathic subjects (8 males, 5 females) ages between 7 to 55 years, (mean age±standard deviation (S.D.): 25.1±17.2 years) We used the dataset which is recorded, preprocessedand features are extracted by Subaşı (2006) for his research namely “Classification of EMG Signals Using Combined Features and Soft computing” in this study. 359 �10.2.Data set The dataset has 129 features which were extracted by AR model from recorded EMG signals and contains three classes which are “Normal”, “Neuropathy” and “Myopathy”. As shown in the table-2 Class Normal Neuropathy Myopathy Total Number of instance 400 399 400 1199 Table-2 11.WEKA WEKA is open source software issued under the GNU General Public License which contains machine learning algorithms for data mining tasks. It is developed for contributing to a theoretical framework for the field by Machine Learning Group at University of Waikato, New Zealand. It composed of easy to use tools which can be applied directly to the dataset. Data pre-processing, classification, regression, clustering, association rules, and visualization are tools in WEKA. And also, well known classification algorithms such as Neural Network, Bayesian, SVM and Decision Tree are available in this tool. It can either get the data from a database or a file. The file format “.arff” and “.cvs” are supported by WEKA. [32] 11.1.Experiments The data mining tool WEKA was used for both feature selection and classification tasks.10fold-Cross validation method was used to train and test the classifiers. In 10-fold crossvalidation, the original sample is randomly partitioned into 10 subsamples. One of the subsample is reserved as the validation data for testing the model, and the residual 9 subsamples are used as training data. The cross-validation process is then repeated 10 times, with each of the 10 subsamples used exactly once as the validation and training data. [33] The data set weretested by four Decision Tree algorithms which are C4.5, Random Tree, Random Forest and Simple CART and the results wererecorded on a table. Then, the Feature Selection methods wereapplied to the data set to determine non effective or comparably less effective features and ineffectual featureswereremoved from data set. The new data set 360 �wastested by four Decision Tree algorithms and the results wererecorded on a table again. This process wasrepeated whit elevendifferent feature selection methods which are listed below. Totally, 48 different testswere done for this study and the total accuracy of each test wasrecorded on a table (Table-3). The tested Feature Selection methods:  Information Gain,  One-R Attribute Evaluator  Chi Squared Attribute  Principal Components Evaluator  Filtered Attribute Evaluator  Relief Attribute Evaluator  Consistency Subset Evaluator  SVM Attribute Evaluator  Filtered Subset Evaluator  Symmetrical uncertainty Attribute Evaluator  Gain Ratio Attribute Evaluator Min Max Average Evaluator Symmetrical uncertainty Attribute SVM Attribute Evaluator Relief F Attribute Evaluator Principal Components One R Attribute Evaluator Gain Ratio Attribute Evaluator Filtered Subset Evaluator Consistency Subset Evaluator Filtered Attribute Evaluator Chi Squared Attribute Evaluator Information Gain All Features (No Reduction) 11.2.Results j48,( C4.5): 96,33 96,25 96,58 96,16 96,25 97,08 96,41 96,25 91,99 96,50 96,41 96,33 96,05 97,08 91,99 Random Forest 98,50 98,67 98,83 99,17 99,25 98,67 98,92 98,92 93,58 98,83 98,50 99,00 98,40 99,25 93,58 Random Tree 96,66 97,16 95,50 97,33 96,00 97,50 97,08 96,91 89,82 97,25 96,75 96,25 96,18 97,50 89,82 Simple CART 96,50 96,41 96,50 96,41 96,58 96,66 96,50 96,41 91,41 96,41 96,58 96,50 96,07 96,66 91,41 Average 97,00 97,12 96,85 97,27 97,02 97,48 97,23 97,12 91,70 97,25 97,06 97,02 96,68 97,62 91,70 Max 98,50 98,67 98,83 99,17 99,25 98,67 98,92 98,92 93,58 98,83 98,50 99,00 98,40 99,25 93,58 Min 96,33 96,25 95,50 96,16 96,00 96,66 96,41 96,25 89,82 96,41 96,41 96,25 96,05 96,66 89,82 361 �Table-3 The accuracy of the classifier varies from %89.82 to %99.25. The most successful algorithm is Random Forest which can classify the data whit %99.25 accuracy by using feature selection method “Consistency Subset Evaluator”.The Classification algorithms C4.5, Random Tree and Simple CART classify the data with the similar accuracy, between %96.33 and %96.50. Reducing the features by using feature selection methods does not considerably affect the accuracy of the classification algorithms accept the “Principal Component”. Principal Component decreases the classification success of all the algorithms which we test.Using feature selection “Filtered Subset Evaluator” increases the success of C4.5, Random Tree and Simple CART classification algorithms, but not considerable, less then %1. Statistics information for Random Forest with the feature selection method Consistency Subset Evaluator Correctly Classified Instances: 1190- 99.2494 % Incorrectly Classified Instances:9 - 0.7506 % Kappa statistic :0.9887 Mean absolute error : 0.023 Root means squared error : 0.087 Relative absolute error : 5.1793 % Root relative squared error : 18.4468 % Total Number of Instances : 1199 362 �Detailed Accuracy by Class TP FP Rate Precision Recall F-Measure Rate Weighted Average ROC Class Area 0.990 0.005 0.990 0.990 0.990 0.998 Normal 0.995 0.003 0.995 0.995 0.995 0.998 Myopathy 0.992 0.004 0.992 0.992 0.992 0.999 Neuropathy 0.992 0.004 0.992 0.992 0.992 0.999 Confusion Matrix Classified as Normal Myopathy Neuropathy Accuracy 396 2 2 %99.00 Myopathy 1 398 1 %99.50 Neuropathy 3 0 396 %99.25 Normal Confusion Matrix shows that none of Neuropathy classifiedas Myopathy and the Myopathy is classified with the maximum accuracy (%99.50) among the 3 classes. 11.3.Discussion Our study shows that it is possible to implement an accurate automatic diagnostic system to classify the EMG signals as Myopathy, Neuropathy and Normal by using Decision Tree algorithms. All the Decision Tree based classification algorithms which we analyses in this study can be used as classifier for creating such a system, but we recommended using Random Forest, as classifier and “Consistency Subset Evaluator” among feature selection 363 �methods for reducing the features. The performance of this system gives the maximum accuracy (%99.25) among the others.The other Decision Tree based Classifiers C4.5, Random Tree and Simple CART may be used without feature reduction. When the results are compared at the Table-3,feature selection methods enhance the performance less than %1. Among the feature reduction methods, we don’t suggest to use “Principle Component” for selecting effective features, because it noticeably decreases the achievement of all classification methods. It decreases the performance of Random Tree from % 96.66 to %89.82. 12.CONCLUSION This study shows that it is possible to design a high performance automatic diagnostic system by using EMG signals which are taken from 27 different subjects. It is necessary to test our system by using data set which is taken more than 200subjects. REFERENCES [1] Preston, D.C., Shapiro B. (2005). Electromyography and Neuromuscular Disorders: Clinical-electrophysiologic Correlations. Philadelphia, PA:Elsevier. ISBN-13: 978-0-75067492-8 [2] Richfield E. K., Cohen B. A., Albers J. W., Review of quantitative and automated needle electromyographic analyses, IEEE Trans Biomed Eng. 28(7) (1981) 506-514. [3] Subasi A., Yilmaz M., Ozcalik H. R., Classification of EMG signals using wavelet neural network, J. Neuroscience Meth. 156 (2006) 360–367. [4] Hassoun M. H., Wang C., Spitzer A. R., NNERVE: Neural network extraction of repetitive vectors for electromyography—Part I: Algorithm, IEEE Trans. Biomed. Eng. 41 (1994) 1039–1052. [5] CHRISTODOULOU C.I., PATTICHIS C.S., “Unsupervided pattern recognition for the classification of EMG signals”, IEEE Trans Biomed Eng 1999, 46(2):169–78 [6] SCHIZAS C.N., PATTICHIS C.S., “Learning systems in biosignal analysis”, BioSystems 1997, 41:105–25. [7] Subaşı. A, Classification of EMG Signals Using Combined Features and Soft Computing Techniques, Applied Soft Computing, Elsevier 2012, G Model ASOC-1517; No. of Pages 11 [8] Katsis C.D., Goletsis Y., Likas A., Fotiadis D.I., Sarmas I. A novel method for automated EMG decomposition and MUAP classification, 2005 Elsevier B.V Artificial Intelligence in Medicine (2006) 37, 55—64 364 �[9] BOZKURT M. R. Emg İşaretlerinin Modern Öntemlerle Önişlenmesi Sınıflandırılması. T.C. Sakarya Üniversitesi Fen Bilimleri Enstitüsü Doktora Tezi 2007 Ve [10] Begg R., Lai D. T. H., Palaniswami M., Computational Intelligence in Biomedical, EngineeringCRC Press, Boca Raton Florida, 2008 [11] Emly, M., L. Gilmore, and S. Roy (1992). Electromyography. IEEE Potentials 11(2), 25–28. [12] Preedy, V. and T. Peters (2002). Skeletal Muscle: Pathology, Diagnosis and Management of Disease. U.K.: Cambridge University Press. [13] Duda R.O. “Pattern Classification” Richard O., Peter E. Hart and David G. -2nd ed. ISBN: 0-471-05669-3,394, 2001, Stork. John Wiley & Sons Inc. [14] Witten I. H., Frank E., Hall M. A., “Data Mining Practical Machine Learning Tools and Techniques”, Third Edition, 64, 2011 Elsevier Inc. [15] Mitchell T. M. “Machine Learning” 52, 53: McGraw-Hill Science/Engineering/Math; (March 1, 1997) ISBN: 0070428077 [16] Quinlan J.R. Induction of Decision Trees, Machine Learning, Morgan Kaufmann (1), 81-106, 1986 [17]. Quinlan, J. R. C4.5: Programs for Machine Learning. Morgan Kaufmann Publishers, 1993 [18] Wu X., · Kumar V., · Quinlan J. R., · Ghosh J., · Yang Q.,· Motoda H. · McLachlan G. J., · Ng A., · Liu B., · Yu P. S., · Zhou Z., · Steinbach M., · Hand D. J., · Steinberg D.,Top 10 algorithms in data mining SURVEY PAPER 27-28Springer-Verlag London Limited 2007 [19]. Fan,W.; Wang, H.; Yu, P. S., Ma, S., Is random model better? On its accuracy and efficiency. In Proceedings of Third IEEE International Conference on Data Mining (ICDM2003). [20]. Fan W. On the Optimality of Probability Estimation by Random Decision Trees IBM T.J. Watson Research, Hawthorne, New York, 2004 American Association for Artificial Intelligence, weifan@us.ibm.com, (www.aaai.org). [21] BreimanLeo Random Forests inMachine Learning, 45, 5–32, 2001, Statistics Department, University of California, Berkeley, CA 94720 Kluwer Academic Publishers. 2001 [22] Caruana, Rich; Karampatziakis, Nikos; Yessenalina, Ainur (2008). "An empirical evaluation of supervised learning in high dimensions". Proceedings of the 25th International Conference on Machine Learning (ICML). [23] Berthold, Michael R. (2010). Guide to Intelligent Data Analysis. Springer London. 365 �[24]. Breiman L, Friedman JH, Olshen RA, Stone CJ Classification and regression trees. Boca Rota, London New York, Washington DC., Chapman&Hall/CRC, 1984. [25] http://en.wikipedia.org/wiki/AR_model, 15.04.2012 [26] NIST/SEMATECH e-Handbook http://www.itl.nist.gov/div898/handbook/, 12.04.2012. of Statistical Methods, [27] Guyon I., Elisseeff A. An introduction to variable and feature selection. J. Mach Learn Res., 3, 1157–1182. 2003 [28] Liu,H. and Motoda,H. Feature Selection for Knowledge Discovery and Data Mining. Kluwer Academic Publishers, Norwell, MA. 1998 [29] Yu,L. and Liu,H. Efficient feature selection via analysis of relevance and redundancy. J. Mach. Learn. Res., 5, 1205–1224. 2004 [30] Daelemans, Walter, Véronique Hoste, Fien De Meulder and Bart Naudts, `Combined Optimization of Feature Selection and Algorithm Parameter Interaction in Machine Learning of Language.' Proceedings of the 14th European Conference on Machine Learning (ECML2003), Lecture Notes in Computer Science 2837, Springer-Verlag, Cavtat-Dubrovnik, Croatia, 84-95, 2003. [31] Saeys Y., Inza I. aki and Larranaga P. Areview of feature selection techniques in bioinformatics V. 23 P. 2507–2517 BIOINFORMATICS, Advance Access publication 2007 doi:10.1093/bioinformatics/btm344 [32] http://www.cs.waikato.ac.nz/ml/weka/, 12.04.2012 [33] McLachlan G.J., Do K.A. Ambroise, C. Analyzing microarray gene expression data. John Wiley & Sons. 2004. 366 � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Extent The size or duration of the resource. 1185 Title A name given to the resource Classification Of Emg Signals Using Decision Tree Methods Author Author Selami , Keleş Abstract A summary of the resource. Nowadays, Usage of EMG signals are increasing very fast among the Medical Professionals to determine specific disorders. Recent Computational Intelligence studies show that EMG signals can be processed by machine learning methods. The aim of our study is to implement an accurate system to classify EMG signals using decision tree algorithms. We preprocessed the EMG signals and used autoregressive method (AR) for feature selection. Features are reduced by different filtering methods and applied to decision tree classification algorithms, namely Simple CART, C4.5, Random Forest and Random Tree. EMG signals are classified as Myopathy, Neuropathy and Normal. All the data are compared each other on the table try to find out the best classification and feature reduction methods. While tree algorithms classify the data with the accuracy between %89, 82 and %99, 25, feature reduction slightly affects the accuracy of the classification methods. It has been shown that a successful automatic diagnostic system implemented to classify EMG signals by using decision tree algorithms. Furthermore, future reduction may help to increase the accuracy of the system. Keywords: EMG, Neuropathy, Myopathy, Simple CART, C4.5, Random Tree, Random Forest, Feature reduction. Date A point or period of time associated with an event in the lifecycle of the resource 2012-05-31 Keywords Keywords. Conference or Workshop Item PeerReviewed H Social Sciences (General) https://omeka.ibu.edu.ba/files/original/f298859d6080a09ea48e495baf2bdf54.pdf d6824003688b30e79e9aeeb27ba19035 Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Extent The size or duration of the resource. 1209 Title A name given to the resource Classification of Fetal State from the Cardiotocogram Recordings using ANN and Simple Logistic Author Author Hakan, Sahin Subasi, Abdulhamit Abstract A summary of the resource. In this study, we present a comparison of machine learning technics using antepartum cardiotocographs performed by SisPorto 2.0 in predicting newborn outcome. CTG is widely used in pregnancy as a technique of measuring fetal well-being, mainly in pregnancies with increased risk of complications. It is a non-invasive way for checking the fetal conditions in the antepartum period. CTG is a continuous electronic record of the baby’s heart rate acquired via an ultrasound transducer placed on the mother’s abdomen. The information efficiently took out from these recordings can be used to envisage pathological state of the fetus and makes an early intervention possible before there is a permanent damage to the fetus. Using features extracted from the FHR and UC signals, the techniques ANN and Simple Logistic was trained to predict the normal and the pathological state. The dataset which consist of 1831 instances with 21 attributes was tested by using the methods which is mentioned above. The CTG recordings were also categorized 1655 of them as normal and 176 of them as pathological by three expert obstetricians’ consensus. They were showed that ANN and Simple Logistic based methods were able to classify the data as normal and pathological with 98.5% and 98.7% accuracy, respectively. Keywords: Cardiotocogram, CTG, SisPorto, Artificial Neural Network (ANN), Simple Logistic, feotus Date A point or period of time associated with an event in the lifecycle of the resource 2012-05-31 Keywords Keywords. Conference or Workshop Item PeerReviewed H Social Sciences (General),T Technology (General) Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Extent The size or duration of the resource. 907 Title A name given to the resource Classroom Corpus Stylistics, Language Acquisition and Intertextuality – A Work-in-Progress Report Author Author Milojković, Marija Abstract A summary of the resource. Although semantic prosody, introduced to the academic community by Louw (1993), has since been the subject of some debate, a consensus on the existence and power of this linguistic phenomenon (diagnosable through corpus analysis alone) has been reached (McEnery and Hardie (2012). Its existence in other languages has also been confirmed, as well as its implications for language teaching (e.g. the Chinese generally tend to use the verb ‘cause’ positively, in defiance of its negative prosody, see Zhang (2009)). The implications of semantic prosodies for translation have also been considered (Stewart 2009). This paper reports the development of corpus stylistics pedagogy based on the key linguistic phenomena discovered by Louw. Along with semantic prosody, these include relexicalisation (a corpus-accessible feature that all literary devices have in common, see Louw (2008)) and logical semantic prosody – subtext (Louw 2010). The existence of subtext has been proved in Russian (Milojkovic 2011), which points to the possibility of its universality. This initial stage of the project will involve second year students of English, University of Belgrade, who do not have a prior grounding in corpus linguistics or literary stylistics. The use of stylistics terminology relevant to Louw’s theory will be avoided. The students will be asked to analyse short excerpts from English poetic and prose texts using reference corpora (the BNC and the corpus of the 1995 edition of the Times newspaper). After the analysis, they will be given the same texts in their existing Serbian translation and asked if the translations incorporate the stylistic features discovered in the originals. For the purposes of this paper the research questions will be 1) what is the successful methodology of a corpus stylistics pedagogy, 2) what is the effect of corpus stylistics methods on students’ awareness of the nuances of language use and 3) what is the effect of the comparison between the original and the translated text. Answers to these questions will be obtained through a combination of a qualitative survey and the teacher’s observation. Date A point or period of time associated with an event in the lifecycle of the resource 2012-05 Keywords Keywords. Conference or Workshop Item PeerReviewed P Philology. Linguistics Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Extent The size or duration of the resource. 3446 Title A name given to the resource CLASSROOM GROUP WORKS UNDER SCRUTINY: A CASE STUDY AT INTERNATIONAL BURCH UNIVERSITY, IN BOSNIA AND HERZEGOVINA Author Author BASTUG, Harun SALIHAGIC, Selma OZKAN, Ahmet Abstract A summary of the resource. In last couple of decades in contemporary societies, educators and educational institutions have required their students to participate actively and devotedly in their educational process in order to prepare them for the future employment which almost always includes team work and cooperation, Therefore, methods that include students’ cooperation and collaboration within group learning have been used increasingly in all levels of teaching and in all subjects. Nevertheless, it seems that in practice this type of learning is still encountering dependence, passivity, and even anxiety on the part of students. Thus, this paper attempts to provide an insight into students’ perspectives on these issues in the international and multicultural environment of International Burch University. Students were given the opportunity to express their own opinions through the interviews evaluated by taking into consideration key elements of cooperative learning situations (Johnson and Johnson, 1991). Thus, the purpose of this research is to indicate possible shortcomings in the implementation of group work methods applied in practice and to attract students’ attention to their importance by offering possible solutions for their overcoming. Date A point or period of time associated with an event in the lifecycle of the resource 2016 Keywords Keywords. Conference or Workshop Item PeerReviewed PE English https://omeka.ibu.edu.ba/files/original/8875d0627aff9664c2b7ed1cc29be923.docx 7b43ab9abf0fcf86fd2322a3e1d76a30 https://omeka.ibu.edu.ba/files/original/0b4897e80a2a662d4d43ec1c2c86aca4.pdf 979071c23ee9e3fbc12fc2fcaf328eff PDF Text Text Classroom Group Works Under Scrutiny: A Case Study at International Burch University Harun Bastug1 , Selma Salihagic2 and Ahmet Özkan3 ABSTRACT In the last couple of decades in contemporary societies, educators and educational institutions have required their students to participate actively and devotedly in their educational process in order to prepare them for the future employment which almost always includes team work and cooperation, Besides, methods that include students’ cooperation and collaboration within group learning have been used increasingly in all levels of teaching and in all subjects. The present study attempts to provide an insight into students’ perspectives on these issues in the international and multicultural environment of International Burch University. Students were given the opportunity to express their own opinions through the interviews evaluated by taking into consideration key elements of cooperative learning situations. The researchers have aimed to indicate possible shortcomings in the implementation of group work methods applied in practice and to attract students’ attention to their importance by offering possible solutions for their overcoming. � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Extent The size or duration of the resource. 3240 Title A name given to the resource Classroom Group Works Under Scrutiny: A Case Study at International Burch University Author Author BASTUG, Harun SALIHAGIC, Selma OZKAN, Ahmet Abstract A summary of the resource. In the last couple of decades in contemporary societies, educators and educational institutions have required their students to participate actively and devotedly in their educational process in order to prepare them for the future employment which almost always includes team work and cooperation, Besides, methods that include students’ cooperation and collaboration within group learning have been used increasingly in all levels of teaching and in all subjects. The present study attempts to provide an insight into students’ perspectives on these issues in the international and multicultural environment of International Burch University. Students were given the opportunity to express their own opinions through the interviews evaluated by taking into consideration key elements of cooperative learning situations. The researchers have aimed to indicate possible shortcomings in the implementation of group work methods applied in practice and to attract students’ attention to their importance by offering possible solutions for their overcoming. Date A point or period of time associated with an event in the lifecycle of the resource 2016 Keywords Keywords. Article PeerReviewed PE English,PR English literature Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Extent The size or duration of the resource. 3389 Title A name given to the resource CLIL ASSESSMENT Author Author Manić, Danijela Abstract A summary of the resource. This paper introduces assessment in general terms, its main function and methods used in schools, as well as new alternative evaluating methods (self and peer assessment, project assessment and portfolio) that are advised to be used in classroom besides oral and written testing. The issue of assessment in the CLIL (Content and Language Integrated Learning) approach is still being resolved. It is the major area of teacher uncertainty in CLIL context since there are no established assessment practices for combined assessment of content and language. In this teaching approach, dual focus is on simultaneous language and content learning achieved by using the foreign language as the medium of instruction. The main concerns refer to what, who, when and how to assess. Do we assess content or language first? Do we sometimes assess one and not the other? Who assesses it, language teacher or content teacher? What tools can we use for assessment? It is important to bear in mind that assessment begins in the early phase of determining curriculum outcomes. The absence of clear language-specific learning objectives in curriculum makes the process of assessing even harder. Therefore, it is necessary to integrate language outcomes into curriculum, along with contentlearning objectives. This article addresses some studies investigating the effects of various assessment methods in practice. It recommends using alternative assessment measures such as portfolios, performancebased tasks, rubrics, descriptors, etc. It is necessary to do some more research to inform practitioners about the possibilities of integrated assessment of both language and content and to see in what context and on what modules these assessment tools may be used effectively and to offer some applicable framework. Keywords: CLIL, assessment, tasks, rubrics, self and peer assessment, portfolio Date A point or period of time associated with an event in the lifecycle of the resource 2014 Keywords Keywords. Conference or Workshop Item PeerReviewed PE English