Search

Copyright of International Journal of InformaticsTechnologies is the property of Institute of Informatics, Gazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Afyon Kocatepe University Journal of Science & Engineering / Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi is the property of Afyon Kocatepe University, Faculty of Science & Literature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Faculty of Educational Sciences is the property of Ankara University, Faculty of Educational Sciences and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Hyperspectral images contain detailed information for classification. However, these data negatively affect the classification results due to their high size, large data volume and strong correlation between adjacent bands. Classification efficiency and accuracy of hyperspectral images can be improved with an appropriate feature selection method. In this study, the Relief-F feature selection algorithm was preferred due to its features such as being independent of the classification model, not taking into account the assumption of multicollinearity, and being able to process noise values. Salinas-A, Indian Pines and Pavia University datasets were used as experimental data to examine the application effect of the Relief-F algorithm. After the applications, the Support Vector Machine classifier showed higher performance in the Salinas-A and Indian Pines datasets after band selection; It has been observed that the classification accuracy of the Random Forest method is largely preserved. The research results show that the Relief-F algorithm determines the most necessary features in hyperspectral images and the number of bands can be reduced by 60% - 70% with a good classification accuracy. [ABSTRACT FROM AUTHOR]

There is often a trade-off between accuracy and interpretability in Machine Learning (ML) models. As the model becomes more complex, generally the accuracy increases and the interpretability decreases. Interpretable Machine Learning (IML) methods have emerged to provide the interpretability of complex ML models while maintaining accuracy. Thus, accuracy remains constant while determining feature importance. In this study, we aim to compare agnostic IML methods including SHAP and ELI5 with the intrinsic IML methods and Feature Selection (FS) methods in terms of the similarity of attribute selection. Also, we compare agnostic IML models (SHAP, LIME, and ELI5) among each other in terms of similarity of local attribute selection. Experimental studies have been conducted on both general and private datasets to predict company default. According to the obtained results, this study confirms the reliability of agnostic IML methods by demonstrating similarities of up to 86% in the selection of attributes compared to intrinsic IML methods and FS methods. Additionally, certain agnostic IML methods can interpret models for local instances. The findings indicate that agnostic IML models can be applied in complex ML models to offer both global and local interpretability while maintaining high accuracy. [ABSTRACT FROM AUTHOR]

Copyright of Dicle University Journal of Engineering / Dicle Üniversitesi Mühendislik Dergisi is the property of Dicle Universitesi and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Defense Sciences / Savunma Bilmleri Dergisi is the property of Turkish Military Academy Defense Sciences Institute and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Gazi Üniversitesi Fen Bilimleri Dergisi Part C: Tasarım ve Teknoloji is the property of Gazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Polytechnic is the property of Journal of Polytechnic and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Overall equipment effectiveness (OEE) describes production efficiency by combining availability, performance, and quality and is used to evaluate production equipment’s performance. This research’s aim is to investigate the potential of the feature selection techniques and the multiple linear regression method, which is one of the machine learning techniques, in successfully predicting the OEE of the corrugated department of a box factory. In the study, six different planned downtimes and information on seventeen different previously known concepts related to activities to be performed are used as input features. Moreover, backward elimination, forward selection, stepwise selection, correlation-based feature selection (CFS), genetic algorithm, random forest, extra trees, ridge regression, lasso regression, and elastic net feature selection methods are proposed to find the most distinctive feature subset in the dataset. As a result of the analyses performed on the data set consisting of 23 features, 1 output and 1204 working days of information, the elastic net - multiple linear regression model, which selects 19 attributes, gave the best average R2 value compared to other models developed. Occam's razor principle is taken into account since there is not a great difference between the average R2 values obtained. Among the models developed according to the principle, the stepwise selection - multiple linear regression model yielded the best R2 value among those that selected the fewest features.

Copyright of Dicle University Journal of Engineering / Dicle Üniversitesi Mühendislik Dergisi is the property of Dicle Universitesi and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Diabetes is a metabolic disease that is common worldwide. The number of people suffering from diabetes is expected to increase every year around the world. This means a negative impact on both the comfort of life of individuals and the health system. In this respect, it is important to diagnose the disease at an early stage. The high dimensionality of the data used for diagnostic purposes has a negative effect on the cost and time of the calculation. To avoid this, it is important to select the most valuable features for diagnosis. In this study, feature selection was made using Salp Swarm Algorithm, Artificial Bee Colony Algorithm, Whale Optimization Algorithm and Ant Colony Algorithm using the samples in the UCI (UCI Machine Learning Repository) data store. In order to evaluate the selected features, accuracy, sensitivity and specificity parameters were calculated using k-Nearest Neighborhood (KNN), Naive Bayes (NB), Support Vector Machine (SVM) and Artificial Neural Networks (ANN) methods. In the calculations for the probability of having diabetes, an accuracy rate of 99.04% was obtained with the k-Nearest Neighborhood method. [ABSTRACT FROM AUTHOR]

Copyright of Karya Journal of Health Science is the property of Karya Journal of Health Science and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In recent years, many studies have been carried out in the field of artificial intelligence in the literature with the development of equipment. Face recognition algorithms have an important place among these developments. Among the face recognition algorithms, the most successful ones are usually deep learning approaches. Models such as SphereFace, CosFace, ArcFace, and MagFace are important deep learning models in the literature. Despite their success, deep learning models are often computationally costly. Therefore, advanced methods are needed to reduce the computational load for these models. One of the most valid methods for this is to choose the most valuable one among embedding features for face recognition. Thus, cost can be reduced, and accuracy values can be increased even more. In this study, the most valuable of the 512 embedded features in the MagFace model was tried to be obtained by using PSO, GA, SCA, and DE optimization algorithms. As a result, accuracy values of 99.83%, 98.57%, and 98.65% were reached for 193, 252, and 280 features selected in the LFW, CFP, and AGEDB datasets, respectively. [ABSTRACT FROM AUTHOR]

One of the most dangerous diseases in the world is a brain tumor. A brain tumor destroys healthy tissue in the brain and then multiplies abnormally, causing increased internal pressure in the skull. This can lead to death if not diagnosed early. Magnetic Resonance Imaging (MRI) is a diagnostic method that is frequently used in soft tissues and gives successful results. In this study, a brain tumor was automatically detected from MR images. For feature extraction, a pre-trained Convolutional Neural Network (CNN) model named MobilenetV2 was used. Then, the ReliefF algorithm was used for feature selection. The features extracted with MobileNetV2 and the features selected with the ReliefF algorithm are given separately to the classifiers and the system performance is tested. As a result of experimental studies, it was seen that the highest performance was obtained with the combination of MobileNetV2 feature extraction, ReliefF algorithm feature selection, and KNN classifier.

Various pandemics have been recorded in world history until today. The Covid-19 outbreak, which emerged at the end of 2019, has recently been a hot topic in the literature. In this work, extreme learning algorithms are presented as a comparative study for predicting the positive rate for the countries: India, Turkey, Italy, USA and UK. The features to be used in the learning phase are determined with the F-test feature selection method. For each extreme learning approach, results are obtained for each country with the root mean square error evaluation criteria. Accordingly, the radial basis kernel function produces the best estimation results, while the linear kernel function has the highest RMSE. Accordingly, the lowest RMSE value has been obtained for India as 4.17E-03 with the radial basis kernel function based ELM. Also, since Turkey's data contains too many outliers, it has the highest RMSE value (0.015 - 0.035) in linear kernel method among the countries.

Overall equipment effectiveness (OEE) describes production efficiency by combining availability, performance, and quality and is used to evaluate production equipment's performance. This research's aim is to investigate the potential of the feature selection techniques and the multiple linear regression method, which is one of the machine learning techniques, in successfully predicting the OEE of the corrugated department of a box factory. In the study, six different planned downtimes and information on seventeen different previously known concepts related to activities to be performed are used as input features. Moreover, backward elimination, forward selection, stepwise selection, correlation-based feature selection (CFS), genetic algorithm, random forest, extra trees, ridge regression, lasso regression, and elastic net feature selection methods are proposed to find the most distinctive feature subset in the dataset. As a result of the analyses performed on the data set consisting of 23 features, 1 output and 1204 working days of information, the elastic net - multiple linear regression model, which selects 19 attributes, gave the best average R² value compared to other models developed. Occam's razor principle is taken into account since there is not a great difference between the average R² values obtained. Among the models developed according to the principle, the stepwise selection - multiple linear regression model yielded the best R² value among those that selected the fewest features. [ABSTRACT FROM AUTHOR]

In this study, the geometric process (GP) model is considered in order to calculate the debugging and testing costs of a software product. Under the assumption of the GP model, the debugging and testing costs of the software product are obtained depending on the first and second moment functions of the GP. It is observed that the values of the first and second moment functions of the process must be known in order to calculate the debugging and testing costs. At the same time, the calculation of moment functions also depends on both the distribution of the first interarrival time of the GP and the estimates of the model and distribution parameters. In this study, the proposed debugging and testing costs are calculated for the data set containing 136 failure times of a real-time command and control system. For this dataset, it has been shown in previous studies that the GP with gamma distribution can be proposed as a model. Under gamma distribution assumption, the maximum likelihood estimates of the model parameters are obtained. Using the estimates of the model parameters, the first and second moment functions of the GP are calculated with the help of the numerical methods proposed for these functions. Finally, the debugging and testing costs are obtained for the data set. [ABSTRACT FROM AUTHOR]

Copyright of Bingol University Journal of Economics & Administrative Science is the property of Bingol University Journal of Economics & Administrative Science and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Sentiment analysis as a text classification problem is a critical task of extracting subjective information from online text documents. An important problem of text classification is high dimensionality. Dimension reduction is an effective way to improve classification performance in machine learning. Reducing irrelevant features can reduce training time and improve classification accuracy. The performance of different feature selection methods may vary depending on the characteristics of different datasets. In this study, the performance of 6 different filter-based feature selection methods (Correlation-based feature selection, Chi-square, Gain ratio, Information gain, OneR, and Symmetric uncertainty coefficient) were tested and compared on 9 different datasets that are frequently used in sentiment classification. Filter scores were calculated for each feature selection method in all datasets. The obtained filter scores were sorted descendingly. New feature subsets were created and classified by adding features to the previous subset from the feature with the highest filter score to the feature with the lowest filter score. The computational results show that the proposed approach achieves average accuracy rates of 94.34% using the Multinomial Naive Bayes classifier for 9 general sentiment classification datasets. Considering the search space, it can be concluded that this approach can be improved and is competitive with existing approaches. [ABSTRACT FROM AUTHOR]

Copyright of Duzce University Journal of Science & Technology is the property of Duzce University Journal of Science & Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Dokuz Eylul University Muhendislik Faculty of Engineering Journal of Science & Engineering / Dokuz Eylül Üniversitesi Mühendislik Fakültesi Fen ve Mühendislik Dergisi is the property of Dokuz Eylul Universitesi Muhendislik Fakultesi Fen ve Muhendislik Dergisi and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

With the development of internet technologies, there has been a significant increase in textual data. Automatic text classification approaches have become important in order for these textual data to become meaningful. Feature selection and feature weighting have an important place in automatic text classification approaches. In this study, the effect of feature weighting methods on local feature selection methods is examined in detail. Two different weighting methods, three different local feature selection methods, three different criteria datasets, and two classifiers were used in the study. The highest Micro-F1 and Macro-F1 scores were 92.88 and 65.55 for the Reuters-21578 dataset, 99.02 and 98.15 for the 20Newsgroup dataset, and 97.19 and 93.40 for the Enron1 dataset. Experimental results show that better results are obtained with the combination of Odds Ratio (OR) feature selection method, Term Frequency (TF) feature weighting and Support Vector Machine (SVM) classifier. [ABSTRACT FROM AUTHOR]

As stated by the World Health Organization, the occurrence of skin cancer has been increasing in recent years. Between 2 and 3 million non-melanoma skin cancers and at least 132.000 malignant skin cancers occur worldwide each year. Appropriate automatic diagnosis of skin lesions and melanoma recognition can greatly improve the early detection of melanomas. Early diagnosis in skin cancer ensures that patients have the correct diagnosis and treatment. In this study, deep features were extracted from skin lesion images to diagnose whether skin cancer is malignant or not, using cubic-type Support Vector Machine (SVM) classifier and pre-trained Convolutional Neural Network (CNN) based AlexNet and ResNet50 deep architectures, and then combined. Then, effective and distinctive features were selected from these deep features with the ReliefF algorithm. Different classifier algorithms were applied to the combined deep features. Cubic type SVM is used as it gives the best results. In the proposed method, the classification accuracy is 92.41% for the Kaggle dataset and 85.17% for the HAM10000 dataset. In experimental studies, it has been observed that the accuracy score of the proposed model is more successful than other studies. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Engineering & Architectural Faculty of Eskisehir Osmangazi University / Eskişehir Osmangazi Üniversitesi Mühendislik ve Mimarlık Fakültesi Dergisi is the property of Eskisehir Osmangazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Erzincan University Journal of Science & Technology is the property of Erzincan Binali Yildirim Universitesi and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

With the increase in data sizes, researchers needed various methods to make the analysis process easier. It is important to reduce the data size and increase the analysis accuracy. When analyzing data, it is necessary not to deal with unnecessary fields and to produce more accurate results with less input. It is one of the most important first steps in feature selection and data analysis. Various machine learning methods are used for feature selection. Univariate Feature Selection, Recursive Feature Elimination, Tree-Based Feature Selection and Principal Component Analysis methods were used in the study. With these methods, the most important ones among the 13 features in the data set were determined. The most important 6, 5, and 4 attributes were separately input, and the cyber-attack method was predicted with the Random Forest algorithm. When the number of features was reduced to 4, the highest accuracy rate of 97.24% was obtained. It has been concluded that the inclusion of related features in the estimation is important in terms of size and speed in this ratio feature selection. With the results obtained, the importance of feature selection on the data has been demonstrated once again. [ABSTRACT FROM AUTHOR]

Since December 2019, the Covid-19 virus afftected our lives and continues to affect the whole world significantly. The investigistion of the indicators of the Covid-19 virus and vaccination studies are of great interest to overcome the Covid- 19 pandemic based on the World health organization recommendations. In this context, many scientific studies have revealed valuable information for the future of the virus. In this study, estimation of the cOvid-19 cases and Classification of changes in the daily number of cases in Turkey was carried out by using support vector machine and k-nearest neighbor algorithms. The indicators that play a critical role in the estimation of the daily patient number classification have been determined as "positivity rate", "fillation rate", "workplace mobility" and "mobility in parks". It has been observed that the k-nearest neighbor algorithm (84.7%) is the most successful algorithm in the estimation of the daily number of cases when considering the highlighted features. [ABSTRACT FROM AUTHOR]

Nerodejeneratif bir hastalık olan Parkinson, dopamin üreten hücrelerin zamanla azalması sonucunda ortaya çıkar. Bu azalma yaşa bağlı olarak değişir. Dünya nüfusunun yaşlandığı gerçeğine göre bakıldığında bu hastalığın ilerleyen yıllarda daha da artacağı söylenebilir. Parkinson hastalığının tanısı oldukça uzun süreli bir iştir. Kesin bir tanı mekanizması olamamakla birlikte çoğunlukla hasta uzun bir süre takibe alınır ve sonrasında Parkinson hastalığına tanı konulabilir. Bu çalışmada, nörologlara yardımcı bir tanı mekanizması önerilmiştir. Ses verileri yardımıyla Parkinson hastalığına sahip olanlar otomatik olarak tespit edilmiştir. Elde edilen özniteliklere min-max normalizasyon işlemi uygulanıp, karınca koloni algoritması (KKA) ile özniteliklerin seçilmesi işlemi ile tespit başarımlarının arttırılması amaçlanmıştır. Hem normalize edilmiş hem KKA ile seçilmiş özniteliklerin başarımı arttırdığı gösterilmiştir. Destek vektör makinalarının ikinci dereceden fonksiyonları ve KKA ile seçilen 30 adet öznitelik ile %87,5 doğruluk, %89,2 duyarlılık, %85,8 özgüllük ve %89,2 hassaslık ile en yüksek başarım değerleri elde edilmiştir.

Copyright of Dicle University Journal of Engineering / Dicle Üniversitesi Mühendislik Dergisi is the property of Dicle Universitesi and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Engineering & Architectural Faculty of Eskisehir Osmangazi University / Eskişehir Osmangazi Üniversitesi Mühendislik ve Mimarlık Fakültesi Dergisi is the property of Eskisehir Osmangazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Electroencephalography (EEG) is a noninvasive method to record brain activities. Among different EEG recording methods, the recording, while a visual stimulation is shown to the subject, is one of the most popular methods. Recently, steady-state visually-evoked potentials (SSVEP) where visual objects are blinking at a fixed frequency have been commonly-used method in brain-computer interfaces. Although various features extracted from SSVEP records have been used, the use of features from wavelet transform should be preferred due to the nonstationary structure of these signals. In this study, the combination of mother wavelet and classifier, which gives the highest accuracy to determine the stimulating frequency, is examined by applying common wavelet features to inputs of classifiers. Features of energy, variance, and entropy were extracted for well-known five EEG frequency bands using six different mother wavelets. Then, classifier performances of six basic classifiers were compared. This study was run for both each subjects individually and all subjects together. Results showed that (i) ANOVA-based feature selection reduces the performances, (ii) there is no unique combination of classifier and mother wavelet while evaluating each subject individually, (iii) the highest performance was achieved by combination of ensemble learner and Reverse Biorthogonal wavelet while evaluating all subjects together. [ABSTRACT FROM AUTHOR]

The diagnosis of epilepsy from the EEG signals is determined by the visual/manual evaluation performed by the neurologist. This evaluation process is laborious and evaluation results vary according to the experience level of neurologists. Therefore, automated systems that will be created using advanced signal processing techniques are important for diagnosis. In this study, a new feature extraction method is proposed using multiple kernel based one-dimensional center symmetric local binary pattern (1D-CSLBP) to identify epileptic seizures. To strengthen this method, levels have been created and multi-level feature extraction has been carried out. Discrete wavelet transform (DWT) was used to generate the levels and feature extraction was performed using the low pass filter coefficient (L bands) obtained at each level. Neighborhood component analysis (NCA) was used to select the most distinctive features. The obtained features are classified using the nearest neighbors (kNN) algorithm. A high performance method was obtained by using multiple kernel NCA and NCA. The 1D-CSLBP and NCAbased method has reached 100.0% accuracy in A-E, A-D-E, D-E, C-E situations. [ABSTRACT FROM AUTHOR]

Automatic Identification System (AIS) is an observation and analysis system that has become compulsory nowadays due to the risks of maritime transportation such as collision, fire, and spillage of hazardous or polluting substances. In the literature, we can see the applications of basic mathematical models, statistical models and machine learning algorithms using AIS data in order to detect these dangers in advance and to make controlled and safe travel of ships. In this study, AIS data have been evaluated as time series, and accuracy comparisons have been made by being developed different models with Autoregressive Integrated Moving Average, Multilayer Perceptron (MLP) and Deep Recurrent Neural Networks (DRNN) beside traditional route estimation model. In addition, feature selection techniques have been weighted in MLP and RDNN models, and new algorithms have been proposed with these improving. Relief, Pearson's Correlation, Gain Ratio and Information Gain (IG) methods were used to compare the accuracy of the route and collision estimations. In order to be used in these accuracy tests, AIS data related into certain times of Çanakkale Strait and Marmara Sea were used. The results showed that all the approaches were close and high accuracy due to the linear movement of the ships in Çanakkale Strait. On the other hand, it has been observed that the best approach in the Marmara Sea was the improved DRNN with IG due to its irregular structure. [ABSTRACT FROM AUTHOR]

Objective: This study presents a method for optimal selection of gene subsets to enhance the non-clinical diagnostic classification and prediction of colorectal cancer using gene expression level of gene expression profiles obtained with an Affymetrix oligonucleotide array. Material and Method: A Hybrid multiobjective Support vector Machine (SVM) feature selection and classification algorithm was employed to determine the Biomarker gene subsets that are highly statistically and clinically relevant to the 62 (tumour or normal) responses of the gene expression levels. The genes selection was done in two stages with the first stage using the Bayesian t-test to prune the non-informative genes and the second stage employed the multi-objective optimization method that allows sequential addition of genes for optimal determination of the pre-selected gene subsets. The SVM with RBF kernel (SVMREF) was fitted sequentially to select the set of near-optimal genes that are correlated with the response class. Results: The optimally selected gene subset yielded an accuracy of 90.1% on the test data that were never used in the building process of the algorithm. Furthermore, the results obtained from the principal component analysis and the complete linkage hierarchical clustering indicated near-perfect discrimination of the two clinical response groups of the colorectal cancer status of the patients. Conclusion: This work has fully demonstrated that non-clinical colon cancer diagnosis and prediction of patients using their gene signatures from the gene microarray expression data is very possible when the appropriate data mining technique tools are used. [ABSTRACT FROM AUTHOR]

Copyright of International Journal of InformaticsTechnologies is the property of Institute of Informatics, Gazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In many electronic nose applications where gas sensors utilizing for a long time, there is an undesirable drift effect on the sensors, which affects the classification quality negatively. Although the sensor drift is inevitable, it is possible to reduce this effect with the calibration transfer methods. This paper presents a comparison study of various multivariate standardization methods to facilitate an effective calibration way on a comprehensive dataset, which is reachable on-line. In this study, three methods applied: direct standardization (DS) orthogonal signal correction (OSC) and piecewise direct standardization (PDS). In addition, these three methods are applied data, which consisted of selected features. The results have shown that the classification success has increased with multivariate calibration technique applied to the selected features. The results also demonstrate that using the best features in the signal processing part can play an important role for the calibration success. This outcome may lead to a new perspective for the future works. [ABSTRACT FROM AUTHOR]

Copyright of International Journal of InformaticsTechnologies is the property of Institute of Informatics, Gazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Selecting a relevant subset of attributes is one of the most important data preprocessing steps of data mining and machine learning solutions. For the classification task, selection is based on the correlation between an attribute and the class attribute. There are various studies on privacy preserving classification. However, there is no attribute selection solution for such work in the literature. In this study, novel attribute selection methods based on the state of the art solution in statistical database security, known as differential privacy, are proposed. The proposed solutions are implemented with the popular data mining library WEKA and experimental results confirm the positive effects of the proposed solutions on classification accuracy. [ABSTRACT FROM AUTHOR]