Classifying educational resources such as videos and articles can be challenging in low-resource languages due to lack of appropriate tools and sufficient labeled data. To overcome this problem, a crosslingual classification method that utilizes resources created in one high-resource language, such as English, to perform classification in many low-resource languages, is proposed. Data scarcity issue is prevented by transferring information from highresources languages to the low-resources ones. First, word embeddings are extracted using one of the frameworks proposed previously, then classifiers are trained using the highresource language documents. Two versions of the method that use different higher-level composition functions are implemented and compared.
{"title":"Classifying Educational Lectures in Low-Resource Languages","authors":"Gihad N. Sohsah, Onur Güzey, Zaina Tarmanini","doi":"10.1109/ICMLA.2016.0076","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0076","url":null,"abstract":"Classifying educational resources such as videos and articles can be challenging in low-resource languages due to lack of appropriate tools and sufficient labeled data. To overcome this problem, a crosslingual classification method that utilizes resources created in one high-resource language, such as English, to perform classification in many low-resource languages, is proposed. Data scarcity issue is prevented by transferring information from highresources languages to the low-resources ones. First, word embeddings are extracted using one of the frameworks proposed previously, then classifiers are trained using the highresource language documents. Two versions of the method that use different higher-level composition functions are implemented and compared.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130205525","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper proposes a novel non-parametric method to robustly embed conditional and posterior distributions to reproducing Kernel Hilbert space (RKHS). Robust embedding is obtained by the eigenvalue decomposition in the RKHS. By retaining only the leading eigenvectors, the noise in data is methodically disregarded. The non-parametric conditional and posterior distribution embedding obtained by our method can be applied to a wide range of Bayesian inference problems. In this paper, we apply it to heterogeneous face recognition and zero-shot object recognition problems. Experimental validation shows that our method produces better results than the comparative algorithms.
{"title":"Robust Kernel Embedding of Conditional and Posterior Distributions with Applications","authors":"M. Nawaz, Omar Arif","doi":"10.1109/ICMLA.2016.0016","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0016","url":null,"abstract":"This paper proposes a novel non-parametric method to robustly embed conditional and posterior distributions to reproducing Kernel Hilbert space (RKHS). Robust embedding is obtained by the eigenvalue decomposition in the RKHS. By retaining only the leading eigenvectors, the noise in data is methodically disregarded. The non-parametric conditional and posterior distribution embedding obtained by our method can be applied to a wide range of Bayesian inference problems. In this paper, we apply it to heterogeneous face recognition and zero-shot object recognition problems. Experimental validation shows that our method produces better results than the comparative algorithms.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121289774","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Topic models employ the Bag-of-Words (BOW) representation, which break terms into constituent words and treat words as surface strings without assuming predefined knowledge about word meaning. In this paper, we propose the Semantic Concept Latent Dirichlet Allocation (SCLDA) and Semantic Concept Hierarchical Dirichlet Process (SCHDP) based approaches by representing text as meaningful concepts rather than words, using a new model known as Bag-of-Concepts (BOC). We propose new algorithms of applying SCLDA and SCHDP into the Concept Chain Queries (CCQ) problem. The algorithms are focused on discovering new semantic relationships between two concepts across documents where relationships found reveal semantic paths linking two concepts across multiple text units. The experiments demonstrate the search quality has been greatly improved, compared with using other LDA or HDP based approaches.
{"title":"Cross-Document Knowledge Discovery Using Semantic Concept Topic Model","authors":"Xin Li, W. Jin","doi":"10.1109/ICMLA.2016.0026","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0026","url":null,"abstract":"Topic models employ the Bag-of-Words (BOW) representation, which break terms into constituent words and treat words as surface strings without assuming predefined knowledge about word meaning. In this paper, we propose the Semantic Concept Latent Dirichlet Allocation (SCLDA) and Semantic Concept Hierarchical Dirichlet Process (SCHDP) based approaches by representing text as meaningful concepts rather than words, using a new model known as Bag-of-Concepts (BOC). We propose new algorithms of applying SCLDA and SCHDP into the Concept Chain Queries (CCQ) problem. The algorithms are focused on discovering new semantic relationships between two concepts across documents where relationships found reveal semantic paths linking two concepts across multiple text units. The experiments demonstrate the search quality has been greatly improved, compared with using other LDA or HDP based approaches.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116377542","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Power loss minimization plays an important role in the appropriate operation of power networks. Line power loss occurs when the power is transmitted through the lines of a network due to the permittivity of lines medium. Transmission loss may increase the dispatch cost of all of the obtained power flows based on market contracts. Hence, the independent system operators should use loss minimization methods to facilitate the implementation of the contracted power transactions. Loss minimization also will improve the security and stability of power network. In this paper, we present a novel loss minimization scheme based on oblivious network design, referred to as oblivious routing-based power flow method. The method is built on the bottom-up oblivious network routing scheme which offers multiple paths from several sources (generation units) to the specific destinations (electric load demands). Although there is limited information regarding other line flows and the current status of network, the routing scheme mathematically guarantees that the power flow solution is an approximation of the optimal solution with a specific competitiveness ratio. In fact, the Our main focus is on the power flow calculation while optimizing power losses. Compared with the recently developed power flow methods, our approach does not depend on the network topology and its performance for both radial and non-radial networks is accurate. Hence, it is suitable to use the propose approach for large-scale loss minimization while determining the power flows. This paper mainly focuses on the theoretical aspect of the proposed method. As our method is based on a novel concept from computer science discipline, we provide sufficient explanation about the preliminaries of oblivious routing scheme.
{"title":"An Oblivious Routing-Based Power Flow Calculation Method for Loss Minimization of Smart Power Networks: A Theoretical Perspective","authors":"Kianoosh G. Boroojeni, M. Amini, S. S. Iyengar","doi":"10.1109/ICMLA.2016.0113","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0113","url":null,"abstract":"Power loss minimization plays an important role in the appropriate operation of power networks. Line power loss occurs when the power is transmitted through the lines of a network due to the permittivity of lines medium. Transmission loss may increase the dispatch cost of all of the obtained power flows based on market contracts. Hence, the independent system operators should use loss minimization methods to facilitate the implementation of the contracted power transactions. Loss minimization also will improve the security and stability of power network. In this paper, we present a novel loss minimization scheme based on oblivious network design, referred to as oblivious routing-based power flow method. The method is built on the bottom-up oblivious network routing scheme which offers multiple paths from several sources (generation units) to the specific destinations (electric load demands). Although there is limited information regarding other line flows and the current status of network, the routing scheme mathematically guarantees that the power flow solution is an approximation of the optimal solution with a specific competitiveness ratio. In fact, the Our main focus is on the power flow calculation while optimizing power losses. Compared with the recently developed power flow methods, our approach does not depend on the network topology and its performance for both radial and non-radial networks is accurate. Hence, it is suitable to use the propose approach for large-scale loss minimization while determining the power flows. This paper mainly focuses on the theoretical aspect of the proposed method. As our method is based on a novel concept from computer science discipline, we provide sufficient explanation about the preliminaries of oblivious routing scheme.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131885199","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
François G. Meyer, Alexander M. Benison, Zachariah Smith, D. Barth
We describe here the recent results of a multidisciplinary effort to design a biomarker that can actively and continuously decode the progressive changes in neuronal organization leading to epilepsy, a process known as epileptogenesis. Using an animal model of acquired epilepsy, we chronically record hippocampal evoked potentials elicited by an auditory stimulus. Using a set of reduced coordinates, our algorithm can identify universal smooth low-dimensional configurations of the auditory evoked potentials that correspond to distinct stages of epileptogenesis. We use a hidden Markov model to learn the dynamics of the evoked potential, as it evolves along these smooth low-dimensional subsets. We provide experimental evidence that the biomarker is able to exploit subtle changes in the evoked potential to reliably decode the stage of epileptogenesis and predict whether an animal will eventually recover from the injury, or develop spontaneous seizures.
{"title":"Decoding Epileptogenesis in a Reduced State Space","authors":"François G. Meyer, Alexander M. Benison, Zachariah Smith, D. Barth","doi":"10.1109/ICMLA.2016.0033","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0033","url":null,"abstract":"We describe here the recent results of a multidisciplinary effort to design a biomarker that can actively and continuously decode the progressive changes in neuronal organization leading to epilepsy, a process known as epileptogenesis. Using an animal model of acquired epilepsy, we chronically record hippocampal evoked potentials elicited by an auditory stimulus. Using a set of reduced coordinates, our algorithm can identify universal smooth low-dimensional configurations of the auditory evoked potentials that correspond to distinct stages of epileptogenesis. We use a hidden Markov model to learn the dynamics of the evoked potential, as it evolves along these smooth low-dimensional subsets. We provide experimental evidence that the biomarker is able to exploit subtle changes in the evoked potential to reliably decode the stage of epileptogenesis and predict whether an animal will eventually recover from the injury, or develop spontaneous seizures.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134496738","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A novel algorithm is proposed in this study for improving the accuracy and robustness of human biometric identification using electrocardiograms (ECG) from mobile devices. The algorithm combines the advantages of both fiducial and non-fiducial ECG features and implements a fully automated, two-stage cascaded classification system using wavelet analysis coupled with probabilistic random forest machine learning. The proposed algorithm achieves a high identification accuracy of 99.43% for the MIT-BIH Arrhythmia database, 99.98% for the MIT-BIH Normal Sinus Rhythm database, 100% for the ECG data acquired from an ECG sensor integrated into a mobile phone, and 98.79% for the PhysioNet Human-ID database acquired from multiple tests within a 6-month span. These results demonstrate the effectiveness and robustness of the proposed algorithm for biometric identification, hence supporting its practicality in applications such as remote healthcare and cloud data security.
{"title":"ECG Biometric Identification Using Wavelet Analysis Coupled with Probabilistic Random Forest","authors":"Robin Tan, M. Perkowski","doi":"10.1109/ICMLA.2016.0038","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0038","url":null,"abstract":"A novel algorithm is proposed in this study for improving the accuracy and robustness of human biometric identification using electrocardiograms (ECG) from mobile devices. The algorithm combines the advantages of both fiducial and non-fiducial ECG features and implements a fully automated, two-stage cascaded classification system using wavelet analysis coupled with probabilistic random forest machine learning. The proposed algorithm achieves a high identification accuracy of 99.43% for the MIT-BIH Arrhythmia database, 99.98% for the MIT-BIH Normal Sinus Rhythm database, 100% for the ECG data acquired from an ECG sensor integrated into a mobile phone, and 98.79% for the PhysioNet Human-ID database acquired from multiple tests within a 6-month span. These results demonstrate the effectiveness and robustness of the proposed algorithm for biometric identification, hence supporting its practicality in applications such as remote healthcare and cloud data security.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"85 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122960585","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A. Raghunath, K. T. Sreekumar, C. S. Kumar, K. I. Ramachandran
High accuracy fault diagnosis systems are extremely important for effective condition based maintenance (CBM) of rotating machines. In this work, we develop a fault diagnosis system using time and frequency domain statistical features as input to a backend support vector machine (SVM) classifier. We evaluate the performance of the baseline system for speed dependent and speed independent performance. We show how feature mapping and feature normalization can help in enhancing the speed independent performance of machine fault diagnosis systems. We first perform feature mapping using locality constrained linear coding (LLC) which maps the input features to a higher dimensional feature space to be used as input to an SVM classifier (LLC-SVM). It is seen that there is a significant improvement in the speed independent performance of the fault identification system. We obtain an improvement of 11.81% absolute and 10.53% absolute respectively for time and frequency domain LLC-SVM systems compared to the respective baseline systems. We then explore variance normalization considering the speed specific variations as noise to further improve the performance of the fault diagnosis system. We obtain a performance improvement of 8.20% absolute and 6.71% absolute respectively over the time and frequency domain LLC-SVM systems. It may be noted that that the variance normalized LLC-SVM system outperforms.
{"title":"Improving Speed Independent Performance of Fault Diagnosis Systems through Feature Mapping and Normalization","authors":"A. Raghunath, K. T. Sreekumar, C. S. Kumar, K. I. Ramachandran","doi":"10.1109/ICMLA.2016.0136","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0136","url":null,"abstract":"High accuracy fault diagnosis systems are extremely important for effective condition based maintenance (CBM) of rotating machines. In this work, we develop a fault diagnosis system using time and frequency domain statistical features as input to a backend support vector machine (SVM) classifier. We evaluate the performance of the baseline system for speed dependent and speed independent performance. We show how feature mapping and feature normalization can help in enhancing the speed independent performance of machine fault diagnosis systems. We first perform feature mapping using locality constrained linear coding (LLC) which maps the input features to a higher dimensional feature space to be used as input to an SVM classifier (LLC-SVM). It is seen that there is a significant improvement in the speed independent performance of the fault identification system. We obtain an improvement of 11.81% absolute and 10.53% absolute respectively for time and frequency domain LLC-SVM systems compared to the respective baseline systems. We then explore variance normalization considering the speed specific variations as noise to further improve the performance of the fault diagnosis system. We obtain a performance improvement of 8.20% absolute and 6.71% absolute respectively over the time and frequency domain LLC-SVM systems. It may be noted that that the variance normalized LLC-SVM system outperforms.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123983389","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper studies about computational burden of a reference modified PID with a neural network prediction for dc-dc converters. Flexible control methods are required to realize a superior transient response since the converter has a nonlinear behavior. However, the computational burden becomes a problem to implement the control to computation devices. In this paper, the neural network is adopted to improve the transient response of output voltage of the dc-dc converter under the consideration of its computational burden. The neural network computation part has a longer computation period than the PID main control part. It can be possible since the neural network gives more than one predictions which are required for the reference modification for each main control period. Therefore, the reference modification can be adopted on every main control period. From results, it is confirmed that the proposed method can improve the transient response effectively with reducing computational burden of neural network control.
{"title":"A Study on Effects of Different Control Period of Neural Network Based Reference Modified PID Control for DC-DC Converters","authors":"H. Maruta, Hironobu Taniguchi, F. Kurokawa","doi":"10.1109/ICMLA.2016.0081","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0081","url":null,"abstract":"This paper studies about computational burden of a reference modified PID with a neural network prediction for dc-dc converters. Flexible control methods are required to realize a superior transient response since the converter has a nonlinear behavior. However, the computational burden becomes a problem to implement the control to computation devices. In this paper, the neural network is adopted to improve the transient response of output voltage of the dc-dc converter under the consideration of its computational burden. The neural network computation part has a longer computation period than the PID main control part. It can be possible since the neural network gives more than one predictions which are required for the reference modification for each main control period. Therefore, the reference modification can be adopted on every main control period. From results, it is confirmed that the proposed method can improve the transient response effectively with reducing computational burden of neural network control.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129155119","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Solar power penetration has made the site-specific energy ratings an essential necessity for utilities, independent systems operators and regional transmission organizations. Since, it leads to the reliable and efficient energy production with the increased levels of solar power integration. This study concentrates on the partitional clustering analysis of monthly average insolation period data for the 75 provinces in Turkey. Together with the k-means clustering algorithm, we use Pearson Correlation, Cosine, Squared Euclidean and City-Block distance measures for the high-dimensional neighborhood measurement and utilize the silhouette width for validating the achieved clustering results. In consequence of comparing the star glyph plots with the k-means clustering results, the most productive and the most unfavorable places among all provinces are mined on the basis of monthly average insolation period.
{"title":"k-Means Partition of Monthly Average Insolation Period Data for Turkey","authors":"M. Yesilbudak, I. Colak, R. Bayindir","doi":"10.1109/ICMLA.2016.0077","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0077","url":null,"abstract":"Solar power penetration has made the site-specific energy ratings an essential necessity for utilities, independent systems operators and regional transmission organizations. Since, it leads to the reliable and efficient energy production with the increased levels of solar power integration. This study concentrates on the partitional clustering analysis of monthly average insolation period data for the 75 provinces in Turkey. Together with the k-means clustering algorithm, we use Pearson Correlation, Cosine, Squared Euclidean and City-Block distance measures for the high-dimensional neighborhood measurement and utilize the silhouette width for validating the achieved clustering results. In consequence of comparing the star glyph plots with the k-means clustering results, the most productive and the most unfavorable places among all provinces are mined on the basis of monthly average insolation period.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"18 5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116694881","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Once a software project has been developed and delivered, any modification to it corresponds to maintenance. Software maintenance (SM) involves modifications to keep a software project usable in a changed or a changing environment, reactive modifications to correct discovered faults, and modifications to improve performance or maintainability. Since the duration of SM should be predicted, in this study, after a statistical analysis of projects maintained on several platforms and programming languages generations, data sets were selected for training and testing multilayer feedforward neural networks (i.e., multilayer perceptron, MLP). These data sets were obtained from the International Software Benchmarking Standards Group. Results based on Wilcoxon statistical tests show that prediction accuracy with the MLP is statistically better than that with the statistical regression models when software projects were maintained on (1) Mid Range platform and coded in programming languages of third generation, and (2) Multi platform and coded in programming languages of fourth generation.
{"title":"Feedforward Neural Networks for Predicting the Duration of Maintained Software Projects","authors":"C. López-Martín","doi":"10.1109/ICMLA.2016.0093","DOIUrl":"https://doi.org/10.1109/ICMLA.2016.0093","url":null,"abstract":"Once a software project has been developed and delivered, any modification to it corresponds to maintenance. Software maintenance (SM) involves modifications to keep a software project usable in a changed or a changing environment, reactive modifications to correct discovered faults, and modifications to improve performance or maintainability. Since the duration of SM should be predicted, in this study, after a statistical analysis of projects maintained on several platforms and programming languages generations, data sets were selected for training and testing multilayer feedforward neural networks (i.e., multilayer perceptron, MLP). These data sets were obtained from the International Software Benchmarking Standards Group. Results based on Wilcoxon statistical tests show that prediction accuracy with the MLP is statistically better than that with the statistical regression models when software projects were maintained on (1) Mid Range platform and coded in programming languages of third generation, and (2) Multi platform and coded in programming languages of fourth generation.","PeriodicalId":356182,"journal":{"name":"2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116983435","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}