Correlation-based Feature Selection Method Research Articles

Performance analysis of different classification algorithms using different feature selection methods on Parkinson's disease detection

BackgroundIn diagnosis of neurodegenerative diseases, the three-dimensional magnetic resonance imaging (3D-MRI) has been heavily researched. Parkinson's disease (PD) is one of the most common neurodegenerative disorders. New methodThe performances of five different classification approaches using five different attribute rankings each followed with an adaptive Fisher stopping criteria feature selection (FS) method are evaluated. To improve the performance of PD detection, a source fusion technique which combines the gray matter (GM) and white (WM) tissue maps and a decision fusion technique which combines the outputs of all classifiers using the correlation-based feature selection (CFS) method by majority voting are used. ResultsAmong the five FS methods, the CFS provides the highest results for all five classification algorithms and the SVM provides the best classification performances for all five different FS methods. The classification accuracy of 77.50% and 81.25% are obtained for the GM and WM tissues, respectively. However, the fusion of GM and WM datasets improves the classification accuracy of the proposed methodology up to 95.00%. Comparison with Existing MethodsAn f-contrast is used to generate 3D masks for GM and WM datasets and a fusion technique, combining the GM and WM datasets is used. Several classification algorithms using several FS methods are performed and a decision fusion technique is used. ConclusionsUsing the combination of the 3D masked GM and WM tissue maps and the fusion of the outputs of multiple classifiers with CFS method gives the classification accuracy of 95.00%.

Novel classification for global gene signature model for predicting severity of systemic sclerosis.

Progression of systemic scleroderma (SSc), a chronic connective tissue disease that causes a fibrotic phenotype, is highly heterogeneous amongst patients and difficult to accurately diagnose. To meet this clinical need, we developed a novel three-layer classification model, which analyses gene expression profiles from SSc skin biopsies to diagnose SSc severity. Two SSc skin biopsy microarray datasets were obtained from Gene Expression Omnibus. The skin scores obtained from the original papers were used to further categorize the data into subgroups of low (<18) and high (≥18) severity. Data was pre-processed for normalization, background correction, centering and scaling. A two-layered cross-validation scheme was employed to objectively evaluate the performance of classification models of unobserved data. Three classification models were used: support vector machine, random forest, and naive Bayes in combination with feature selection methods to improve performance accuracy. For both input datasets, random forest classifier combined with correlation-based feature selection (CFS) method and naive Bayes combined with CFS or support vector machine based recursive feature elimination method yielded the best results. Additionally, we performed a principal component analysis to show that low and high severity groups are readily separable by gene expression signatures. Ultimately, we found that our novel classification prediction model produced global gene signatures that significantly correlated with skin scores. This study represents the first report comparing the performance of various classification prediction models for gene signatures from SSc patients, using current clinical diagnostic factors. In summary, our three-classification model system is a powerful tool for elucidating gene signatures from SSc skin biopsies and can also be used to develop a prognostic gene signature for SSc and other fibrotic disorders.

Risk prediction in life insurance industry using supervised learning algorithms

Risk assessment is a crucial element in the life insurance business to classify the applicants. Companies perform underwriting process to make decisions on applications and to price policies accordingly. With the increase in the amount of data and advances in data analytics, the underwriting process can be automated for faster processing of applications. This research aims at providing solutions to enhance risk assessment among life insurance firms using predictive analytics. The real world dataset with over hundred attributes (anonymized) has been used to conduct the analysis. The dimensionality reduction has been performed to choose prominent attributes that can improve the prediction power of the models. The data dimension has been reduced by feature selection techniques and feature extraction namely, Correlation-Based Feature Selection (CFS) and Principal Components Analysis (PCA). Machine learning algorithms, namely Multiple Linear Regression, Artificial Neural Network, REPTree and Random Tree classifiers were implemented on the dataset to predict the risk level of applicants. Findings revealed that REPTree algorithm showed the highest performance with the lowest mean absolute error (MAE) value of 1.5285 and lowest root-mean-squared error (RMSE) value of 2.027 for the CFS method, whereas Multiple Linear Regression showed the best performance for the PCA with the lowest MAE and RMSE values of 1.6396 and 2.0659, respectively, as compared to the other models.

Causal risk factor discovery for severe acute kidney injury using electronic health records

BackgroundAcute kidney injury (AKI), characterized by abrupt deterioration of renal function, is a common clinical event among hospitalized patients and it is associated with high morbidity and mortality. AKI is defined in three stages with stage-3 being the most severe phase which is irreversible. It is important to effectively discover the true risk factors in order to identify high-risk AKI patients and allow better targeting of tailored interventions. However, Stage-3 AKI patients are very rare (only 0.2% of AKI patients) with a large scale of features available in EHR (1917 potential risk features), yielding a scenario unfeasible for any correlation-based feature selection or modeling method. This study aims to discover the key factors and improve the detection of Stage-3 AKI.MethodsA causal discovery method (McDSL) is adopted for causal discovery to infer true causal relationship between information buried in EHR (such as medication, diagnosis, laboratory tests, comorbidities and etc.) and Stage-3 AKI risk. The research approach comprised two major phases: data collection, and causal discovery. The first phase is propose to collect the data from HER (includes 358 encounters and 891 risk factors). Finally, McDSL is employed to discover the causal risk factors of Stage-3 AKI, and five well-known machine learning models are built for predicting Stage-3 AKI with 10-fold cross-validation (predictive accuracy were measured by AUC, precision, recall and F-score).ResultsMcDSL is useful for further research of EHR. It is able to discover four causal features, all selected features are medications that are modifiable. The latest research of machine learning is employed to compare the performance of prediction, and the experimental result has verified the selected features are pivotal.ConclusionsThe features selected by McDSL, which enable us to achieve significant dimension reduction without sacrificing prediction accuracy, suggesting potential clinical use such as helping physicians develop better prevention and treatment strategies.

Research on social data by means of cluster analysis

This paper presents a data mining study and cluster analysis of social data obtained on small producers and family farmers from six country cities in Ceará state, northeast Brazil. The analyzed data involve demographic, economic, agriculture and food insecurity information. The goal of the study is to establish profiles for the small producer families that reside in the region and to identify relevant features which differentiate these profiles. Moreover, we provide an efficient data mining methodology for analysis of social data sets which is capable of handling its natural challenges, such as mixed variables and abundance of null values. We use the Silhouette method for the estimation of the best number of natural groups within the data, along with the Partitioning Around Medoids clustering algorithm in order to compute the profiles. The Correlation-Based Feature Selection method is used to identify which social criteria are the most important to differentiate the families from each profile. Classification models based on support vector machines, multilayer perceptron and decision trees were developed aiming to predict in which of the identified clusters an arbitrary family would be best fit. We obtained a good separation of the families into two clusters, and a multilayer perceptron model with approximately 93.5% prediction accuracy.

Co-ABC: Correlation artificial bee colony algorithm for biomarker gene discovery using gene expression profile.

In this paper, we propose a new hybrid method based on Correlation-based feature selection method and Artificial Bee Colony algorithm,namely Co-ABC to select a small number of relevant genes for accurate classification of gene expression profile. The Co-ABC consists of three stages which are fully cooperated: The first stage aims to filter noisy and redundant genes in high dimensionality domains by applying Correlation-based feature Selection (CFS) filter method. In the second stage, Artificial Bee Colony (ABC) algorithm is used to select the informative and meaningful genes. In the third stage, we adopt a Support Vector Machine (SVM) algorithm as classifier using the preselected genes form second stage. The overall performance of our proposed Co-ABC algorithm was evaluated using six gene expression profile for binary and multi-class cancer datasets. In addition, in order to proof the efficiency of our proposed Co-ABC algorithm, we compare it with previously known related methods. Two of these methods was re-implemented for the sake of a fair comparison using the same parameters. These two methods are: Co-GA, which is CFS combined with a genetic algorithm GA. The second one named Co-PSO, which is CFS combined with a particle swarm optimization algorithm PSO. The experimental results shows that the proposed Co-ABC algorithm acquire the accurate classification performance using small number of predictive genes. This proofs that Co-ABC is a efficient approach for biomarker gene discovery using cancer gene expression profile.

Comparison of Feature Selection Methods to Classify Inhibitors in DUD-E Database

In designing a new drug, inhibitor compound is usually used to control the enzyme work to recover a particular disease. In the drug design technique, the classification of inhibitor is carry out by docking software to simulate the bounding of mixing (new inhibitor candidate) with the targeted enzyme. DUD-E is a database to simulate docking with high dimensional data characteristic, which lead to the feasibility of machine learning approach as the analytical tool. A compound with specific characterictics can be classified into ligand or decoy by using many characterictics leading to a problem in the machine learning algorithm. This paper discusses feature selection analysis to obtain the compound characteristics which are effectively determine ligand or decoy. This paper examined Mutual Information-based Feature Selection (MIFS), Correlation-based Feature Selection (CFS) as well as Fast Correlation-Based Filter (FCBF), and the results show that the FCBF always selects less number of features with fastest runtime of classification. The highest classification accuracy is obtained when all features are used in the classification by k-NN. However, the accuracy is slightly different with classification using selected features. The CFS method performs well for Data-A with accuracy of 89,55%, while the MIFS outperforms the others for Data-B and Data-C with the classification accuracy of 92,34% and 95,20% consecutively.

An efficient categorization of liver cirrhosis using convolution neural networks for health informatics

Accurate categorization of cirrhosis liver image in ultrasound modality is of great importance in medical diagnosis and treatment. Health informatics is important as it provides quick predictions on diseases based on nearness of symptoms. Modelling solutions for the same on cloud using deep learning is the motivation of this paper. Here, we propose a deep learning model associated with correlation based feature selection method for cirrhosis image classification. We compare the results with three other conventional classifiers algorithms to improve the better classification accuracy. First in pre-processing stage, noises are eliminated from pathological scan images by using modified laplacian pyramid non-linear diffusion filter. From the pre-processed scan images, each cirrhosis region is obtained under the guidance of radiology or physicians. Then, after extracting the complete features of each patch by gray level, local binary pattern and scale invariant feature, a feature selection technique is applied to choice the predominant texture features for each classifier. Finally a convolution neural network is implemented to improve the performance of classifiers in terms of sensitivity, specificity and accuracy. Convolution neural network algorithm with two hidden layers gives more accuracy in classifying cirrhosis image with 98% sensitivity. Experiments are carried out with 990 cirrhosis image patches which demonstrates that our proposed deep learning classifier perform 100% well than original classifiers in terms of accuracy.

Cancer Classification Based on Support Vector Machine Optimized by Particle Swarm Optimization and Artificial Bee Colony

Intelligent optimization algorithms have advantages in dealing with complex nonlinear problems accompanied by good flexibility and adaptability. In this paper, the FCBF (Fast Correlation-Based Feature selection) method is used to filter irrelevant and redundant features in order to improve the quality of cancer classification. Then, we perform classification based on SVM (Support Vector Machine) optimized by PSO (Particle Swarm Optimization) combined with ABC (Artificial Bee Colony) approaches, which is represented as PA-SVM. The proposed PA-SVM method is applied to nine cancer datasets, including five datasets of outcome prediction and a protein dataset of ovarian cancer. By comparison with other classification methods, the results demonstrate the effectiveness and the robustness of the proposed PA-SVM method in handling various types of data for cancer classification.

Learning Lung Nodule Malignancy Likelihood from Radiologist Annotations or Diagnosis Data

Lung cancer is the world’s most lethal type of cancer, being crucial that an early diagnosis is made in order to achieve successful treatments. Computer-aided diagnosis can play an important role in lung nodule detection and on establishing the nodule malignancy likelihood. This paper is a contribution in the design of a learning approach, using computed tomography images. Our methodology involves the measurement of a set of features in the nodular image region, and train classifiers, as K-nearest neighbor or support vector machine (SVM), to compute the malignancy likelihood of lung nodules. For this purpose, the Lung Image Database Consortium and image database resource initiative database is used due to its size and nodule variability, as well as for being publicly available. For training we used both radiologist’s labels and annotations and diagnosis data, as biopsy, surgery and follow-up results. We obtained promising results, as an Area Under the Receiver operating characteristic curve value of 0.962 ± 0.005 and 0.905 ± 0.04 was achieved for the Radiologists’ data and for the Diagnosis data, respectively, using an SVM with an exponential kernel combined with a correlation-based feature selection method.

Predicting CD4 count changes among patients on antiretroviral treatment: Application of data mining techniques

Background and objectivesTo monitor the progress of therapy and disease progression, periodic CD4 counts are required throughout the course of HIV/AIDS care and support. The demand for CD4 count measurement is increasing as ART programs expand over the last decade. This study aimed to predict CD4 count changes and to identify the predictors of CD4 count changes among patients on ART. MethodsA cross-sectional study was conducted at the University of Gondar Hospital from 3,104 adult patients on ART with CD4 counts measured at least twice (baseline and most recent). Data were retrieved from the HIV care clinic electronic database and patients` charts. Descriptive data were analyzed by SPSS version 20. Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology was followed to undertake the study. WEKA version 3.8 was used to conduct a predictive data mining. Before building the predictive data mining models, information gain values and correlation-based Feature Selection methods were used for attribute selection. Variables were ranked according to their relevance based on their information gain values. J48, Neural Network, and Random Forest algorithms were experimented to assess model accuracies. ResultThe median duration of ART was 191.5 weeks. The mean CD4 count change was 243 (SD 191.14) cells per microliter. Overall, 2427 (78.2%) patients had their CD4 counts increased by at least 100 cells per microliter, while 4% had a decline from the baseline CD4 value. Baseline variables including age, educational status, CD8 count, ART regimen, and hemoglobin levels predicted CD4 count changes with predictive accuracies of J48, Neural Network, and Random Forest being 87.1%, 83.5%, and 99.8%, respectively. Random Forest algorithm had a superior performance accuracy level than both J48 and Artificial Neural Network. The precision, sensitivity and recall values of Random Forest were also more than 99%. ConclusionsNearly accurate prediction results were obtained using Random Forest algorithm. This algorithm could be used in a low-resource setting to build a web-based prediction model for CD4 count changes.

Apple leaf disease identification using genetic algorithm and correlation based feature selection method

Apple leaf disease is one of the main factors to constrain the apple production and quality. It takes a long time to detect the diseases by using the traditional diagnostic approach, thus farmers often miss the best time to prevent and treat the diseases. Apple leaf disease recognition based on leaf image is an essential research topic in the field of computer vision, where the key task is to find an effective way to represent the diseased leaf images. In this research, based on image processing techniques and pattern recognition methods, an apple leaf disease recognition method was proposed. A color transformation structure for the input RGB (Red, Green and Blue) image was designed firstly and then RGB model was converted to HSI (Hue, Saturation and Intensity), YUV and gray models. The background was removed based on a specific threshold value, and then the disease spot image was segmented with region growing algorithm (RGA). Thirty-eight classifying features of color, texture and shape were extracted from each spot image. To reduce the dimensionality of the feature space and improve the accuracy of the apple leaf disease identification, the most valuable features were selected by combining genetic algorithm (GA) and correlation based feature selection (CFS). Finally, the diseases were recognized by SVM classifier. In the proposed method, the selected feature subset was globally optimum. The experimental results of more than 90% correct identification rate on the apple diseased leaf image database which contains 90 disease images for there kinds of apple leaf diseases, powdery mildew, mosaic and rust, demonstrate that the proposed method is feasible and effective. Keywords: apple leaf disease, diseased leaf recognition, region growing algorithm (RGA), genetic algorithm and correlation based feature selection (GA-CFS) DOI: 10.3965/j.ijabe.20171002.2166

Automated epileptic seizure detection using improved correlation-based feature selection with random forest classifier

Analysis of electroencephalogram (EEG) signal is crucial due to its non-stationary characteristics, which could lead the way to proper detection method for the treatment of patients with neurological abnormalities, especially for epilepsy. The performance of EEG-based epileptic seizure detection relies largely on the quality of selected features from an EEG data that characterize seizure activity. This paper presents a novel analysis method for detecting epileptic seizure from EEG signal using Improved Correlation-based Feature Selection method (ICFS) with Random Forest classifier (RF). The analysis involves, first applying ICFS to select the most prominent features from the time domain, frequency domain, and entropy based features. An ensemble of Random Forest (RF) classifiers is then learned on the selected set of features. The experimental results demonstrate that the proposed method shows better performance compared to the conventional Correlation-based method and also outperforms some other state-of-the-art methods of epileptic seizure detection using the same benchmark EEG dataset.

DoS Detection Method based on Artificial Neural Networks

DoS attack tools have become increasingly sophis-ticated challenging the existing detection systems to continually improve their performances. In this paper we present a victim-end DoS detection method based on Artificial Neural Networks (ANN). In the proposed method a Feed-forward Neural Network (FNN) is optimized to accurately detect DoS attack with minimum resources usage. The proposed method consists of the following three major steps: (1) Collection of the incoming network traffic,(2) selection of relevant features for DoS detection using an unsupervised Correlation-based Feature Selection (CFS) method,(3) classification of the incoming network traffic into DoS traffic or normal traffic. Various experiments were conducted to evaluate the performance of the proposed method using two public datasets namely UNSW-NB15 and NSL-KDD. The obtained results are satisfactory when compared to the state-of-the-art DoS detection methods.

PREDICTIVE MODELING OF HOSPITAL READMISSION RATES USING ELECTRONIC MEDICAL RECORD-WIDE MACHINE LEARNING: A CASE-STUDY USING MOUNT SINAI HEART FAILURE COHORT.

Reduction of preventable hospital readmissions that result from chronic or acute conditions like stroke, heart failure, myocardial infarction and pneumonia remains a significant challenge for improving the outcomes and decreasing the cost of healthcare delivery in the United States. Patient readmission rates are relatively high for conditions like heart failure (HF) despite the implementation of high-quality healthcare delivery operation guidelines created by regulatory authorities. Multiple predictive models are currently available to evaluate potential 30-day readmission rates of patients. Most of these models are hypothesis driven and repetitively assess the predictive abilities of the same set of biomarkers as predictive features. In this manuscript, we discuss our attempt to develop a data-driven, electronic-medical record-wide (EMR-wide) feature selection approach and subsequent machine learning to predict readmission probabilities. We have assessed a large repertoire of variables from electronic medical records of heart failure patients in a single center. The cohort included 1,068 patients with 178 patients were readmitted within a 30-day interval (16.66% readmission rate). A total of 4,205 variables were extracted from EMR including diagnosis codes (n=1,763), medications (n=1,028), laboratory measurements (n=846), surgical procedures (n=564) and vital signs (n=4). We designed a multistep modeling strategy using the Naïve Bayes algorithm. In the first step, we created individual models to classify the cases (readmitted) and controls (non-readmitted). In the second step, features contributing to predictive risk from independent models were combined into a composite model using a correlation-based feature selection (CFS) method. All models were trained and tested using a 5-fold cross-validation method, with 70% of the cohort used for training and the remaining 30% for testing. Compared to existing predictive models for HF readmission rates (AUCs in the range of 0.6-0.7), results from our EMR-wide predictive model (AUC=0.78; Accuracy=83.19%) and phenome-wide feature selection strategies are encouraging and reveal the utility of such datadriven machine learning. Fine tuning of the model, replication using multi-center cohorts and prospective clinical trial to evaluate the clinical utility would help the adoption of the model as a clinical decision system for evaluating readmission status.

A two-stage method for microcalcification cluster segmentation in mammography by deformable models.

Segmentation of microcalcification (MC) clusters in x-ray mammography is a difficult task for radiologists. Accurate segmentation is prerequisite for quantitative image analysis of MC clusters and subsequent feature extraction and classification in computer-aided diagnosis schemes. In this study, a two-stage semiautomated segmentation method of MC clusters is investigated. The first stage is targeted to accurate and time efficient segmentation of the majority of the particles of a MC cluster, by means of a level set method. The second stage is targeted to shape refinement of selected individual MCs, by means of an active contour model. Both methods are applied in the framework of a rich scale-space representation, provided by the wavelet transform at integer scales. Segmentation reliability of the proposed method in terms of inter and intraobserver agreements was evaluated in a case sample of 80 MC clusters originating from the digital database for screening mammography, corresponding to 4 morphology types (punctate: 22, fine linear branching: 16, pleomorphic: 18, and amorphous: 24) of MC clusters, assessing radiologists' segmentations quantitatively by two distance metrics (Hausdorff distance-HDISTcluster, average of minimum distance-AMINDISTcluster) and the area overlap measure (AOMcluster). The effect of the proposed segmentation method on MC cluster characterization accuracy was evaluated in a case sample of 162 pleomorphic MC clusters (72 malignant and 90 benign). Ten MC cluster features, targeted to capture morphologic properties of individual MCs in a cluster (area, major length, perimeter, compactness, and spread), were extracted and a correlation-based feature selection method yielded a feature subset to feed in a support vector machine classifier. Classification performance of the MC cluster features was estimated by means of the area under receiver operating characteristic curve (Az ± Standard Error) utilizing tenfold cross-validation methodology. A previously developed B-spline active rays segmentation method was also considered for comparison purposes. Interobserver and intraobserver segmentation agreements (median and [25%, 75%] quartile range) were substantial with respect to the distance metrics HDISTcluster (2.3 [1.8, 2.9] and 2.5 [2.1, 3.2] pixels) and AMINDISTcluster (0.8 [0.6, 1.0] and 1.0 [0.8, 1.2] pixels), while moderate with respect to AOMcluster (0.64 [0.55, 0.71] and 0.59 [0.52, 0.66]). The proposed segmentation method outperformed (0.80 ± 0.04) statistically significantly (Mann-Whitney U-test, p < 0.05) the B-spline active rays segmentation method (0.69 ± 0.04), suggesting the significance of the proposed semiautomated method. Results indicate a reliable semiautomated segmentation method for MC clusters offered by deformable models, which could be utilized in MC cluster quantitative image analysis.

Detection of abnormalities in lumbar discs from clinical lumbar MRI with hybrid models

Disc abnormalities cause a great number of complaints including lower back pain. Lower back pain is one of the most common types of pain in the world. The computer-assisted detection of this ailment will be of great use to physicians and specialists. With this study, hybrid models have been developed which include feature extraction, selection and classification characteristics for the purpose of determining the disc abnormalities in the lumbar region. In determining the abnormalities, T2-weighted sagittal and axial Magnetic Resonance Images (MRI) were taken from 55 people. In the feature extraction stage, 27 appearance characteristics and form characteristics were acquired from both sagittal and transverse images. In the feature selection stage, the F-Score-Based Feature Selection (FSFS) and the Correlation-Based Feature Selection (CBFS) methods were used to select the best discriminative features. The number of features was reduced to 5 from 27 by using the FSFS, and to 22 from 27 by using the CBFS. In the last stage, five different classification algorithms, i.e. the Multi-Layer Perceptron, the Support Vector Machine, the Decision Tree, the Naïve Bayes, and the k Nearest Neighbor algorithms were applied. In addition, the combination of the classifier model (the combination of the bagging and the random forests) has been used to improve the classification performance in the detection of lumbar disc datasets. The results which were obtained suggest that the proposed hybrid models can be used safely in detecting the disc abnormalities.

A New Direction of Cancer Classification: Positive Effect of Low-Ranking MicroRNAs

ObjectivesMany studies based on microRNA (miRNA) expression profiles showed a new aspect of cancer classification. Because one characteristic of miRNA expression data is the high dimensionality, feature selection methods have been used to facilitate dimensionality reduction. The feature selection methods have one shortcoming thus far: they just consider the problem of where feature to class is 1:1 or n:1. However, because one miRNA may influence more than one type of cancer, human miRNA is considered to be ranked low in traditional feature selection methods and are removed most of the time. In view of the limitation of the miRNA number, low-ranking miRNAs are also important to cancer classification. MethodsWe considered both high- and low-ranking features to cover all problems (1:1, n:1, 1:n, and m:n) in cancer classification. First, we used the correlation-based feature selection method to select the high-ranking miRNAs, and chose the support vector machine, Bayes network, decision tree, k-nearest-neighbor, and logistic classifier to construct cancer classification. Then, we chose Chi-square test, information gain, gain ratio, and Pearson's correlation feature selection methods to build the m:n feature subset, and used the selected miRNAs to determine cancer classification. ResultsThe low-ranking miRNA expression profiles achieved higher classification accuracy compared with just using high-ranking miRNAs in traditional feature selection methods. ConclusionOur results demonstrate that the m:n feature subset made a positive impression of low-ranking miRNAs in cancer classification.

Automatic detection of overnight deep sleep based on heart rate variability: a preliminary study.

This preliminary study investigated the use of cardiac information or more specifically, heart rate variability (HRV), for automatic deep sleep detection throughout the night. The HRV data can be derived from cardiac signals, which were obtained from polysomnography (PSG) recordings. In total 42 features were extracted from the HRV data of 15 single-night PSG recordings (from 15 healthy subjects) for each 30-s epoch, used to perform epoch-by-epoch classification of deep sleep and non-deep sleep (including wake state and all the other sleep stages except deep sleep). To reduce variation of cardiac physiology between subjects, we normalized each feature per subject using a simple Z-score normalization method by subtracting the mean and dividing by the standard deviation of the feature values. A correlation-based feature selection (CFS) method was employed to select informative features as well as removing feature redundancy and a linear discriminant (LD) classifier was applied for deep and non-deep sleep classification. Results show that the use of Z-score normalization can significantly improve the classification performance. A Cohen's Kappa coefficient of 0.42 and an overall accuracy of 81.3% based on a leave-one-subject-out cross-validation were achieved.

An intelligent system for lung cancer diagnosis using a new genetic algorithm based feature selection method.

In this paper, we develop a novel feature selection algorithm based on the genetic algorithm (GA) using a specifically devised trace-based separability criterion. According to the scores of class separability and variable separability, this criterion measures the significance of feature subset, independent of any specific classification. In addition, a mutual information matrix between variables is used as features for classification, and no prior knowledge about the cardinality of feature subset is required. Experiments are performed by using a standard lung cancer dataset. The obtained solutions are verified with three different classifiers, including the support vector machine (SVM), the back-propagation neural network (BPNN), and the K-nearest neighbor (KNN), and compared with those obtained by the whole feature set, the F-score and the correlation-based feature selection methods. The comparison results show that the proposed intelligent system has a good diagnosis performance and can be used as a promising tool for lung cancer diagnosis.