FEATURE SELECTION TECHNIQUES

The course selection process in higher education has grown more complex as the students face the growing volume of academic options that must align with their personal and career goals. The traditional advising methods often fail to provide the personalized recommendations, resulting in the delayed graduations and the higher dropout rates. Therefore, the automated and intelligent course selection models and systems are becoming more important. These AI-based systems nearly deliver the tailored guidance to the students by analyzing their academic performance. In addition, this guidance is also based on the students' individual characteristics, helping to create the personalized course recommendations. This study presents a predictive model for the personalized course advising, in which it incorporates data from the academic, personal, and the social dimensions. Then, this model uses and employs the advanced machine learning algorithms and the optimization techniques and methods to maximize the accuracy of course recommendations. The feature selection methods like Information Gain and Chi-Square are used, in addition to dedicated classifiers such as the Decision Trees (DT) and the K-Nearest Neighbors (KNN). To balance the dataset, the Synthetic Minority Oversampling Technique (SMOTE) is applied. In the implementation phase, first, the Knapsack Problem algorithm is applied to optimize the students' course loads in order to ensure the recommendations match their abilities and the institutional requirements. Next, the KNN is applied and reaches an accuracy of 91.57% and outperforms the DT, which provides a practical solution for the problem of adaptive course advising. INDEX TERMS Personalized academic advising, educational data mining (EDM), feature selection techniques, knapsack problem algorithm.

In today’s world, social media has spread widely, and the social life of people have become deeply associated with social media use. They use it to communicate with each other, share events and news, and even run businesses. The huge growth in social media and the massive number of users has lured attackers to distribute harmful content through fake accounts, leading to a large number of people falling victim to those accounts. In this work, we propose a mechanism for identifying fake accounts on the social media site Twitter by using two methods to preprocess data and extract the most effective features, they are the spearman correlation coefficient and the chi-square test. For classification, we used supervised machine learning algorithms based on the ensemble system (stack method) by using random forest, support vector machine, and Naive Bayes algorithms in the first level of the stack, and the logistic regression algorithm as a meta classifier. The stack ensemble system was show...

Proteomic analysis of cancers' stages has provided new opportunities for the development of novel, highly sensitive diagnostic tools which helps early detection of cancer. This paper introduces a new feature ranking approach called FRMT. FRMT is based on the Technique for Order of Preference by Similarity to Ideal Solution method (TOPSIS) which select the most discriminative proteins from proteomics data for cancer staging. In this approach, outcomes of 10 feature selection techniques were combined by TOPSIS method, to select the final discriminative proteins from seven different proteomic databases of protein expression profiles. In the proposed workflow, feature selection methods and protein expressions have been considered as criteria and alternatives in TOPSIS, respectively. The proposed method is tested on seven various classifier models in a 10-fold cross validation procedure that repeated 30 times on the seven cancer datasets. The obtained results proved the higher stability and superior classification performance of method in comparison with other methods, and it is less sensitive to the applied classifier. Moreover, the final introduced proteins are informative and have the potential for application in the real medical practice.

Nowadays, fake news is prevalent and too simple to propagate through social media, particularly during elections and pandemics like COVID-19. Several fake news stories have appeared on social media sites like LINE, Facebook, and Twitter after the COVID-19 epidemic throughout the world. Also, a lot of older individuals simply forward these communications without checking their veracity, which speeds up the dissemination of fake information. So, our goal is to identify fake news using machine learning. In this paper, we describe a supervised method that automatically gathers a sizable but noisy training dataset made up of a significant number of tweets. We will categorize tweets during collection into trustworthy and untrustworthy sources, then using the dataset to train a classifier. The categorization of fake and real tweets is the next classification objective for which we apply that classifier. We first demonstrate that real news is larger in size, shared on Twitter for a longer length of time, and shared by people with more followers than following. Second, we employed machine learning models like support vector machine (SVM), random forest (RF), and decision tree (DT), and we found out that the SVM is the best of all the models due to its best results and 99% accuracy.

Antimicrobial peptides (AMP) are a promising source of antibiotics with a broad spectrum activity against bacteria and low incidence of developing resistance. The mechanism by which an AMP executes its function depends on a set of computable physicochemical properties from the amino acid sequence. Peptides package was designed for allowing the quick and easy computation of ten structural characteristics own of the antimicrobial peptides, with the aim of generating data for increase the accuracy in classification and design of new amino acid sequences. Moreover, the options to read and plot XVG output files from GROMACS molecular dynamics package was included. Peptides package requires R version 1.2.2 or higher. Development releases of the package are available on the GitHub repository http://github.com/dosorio/peptides. > aacomp(seq = "GLPRKILCAIAKKKGKCKGPLKLVCKC"

Research on the classification of the different categories of small peptides is becoming a challenge for bioinformatics domain. However, machine learning models have shown their potential to tackle such applications. We propose a multi-class random forest (MCRF) classifier to classify small peptides which is compared with state-ofart classifiers including, support vector machine with RBF kernel (SVM + RBF), naïve Bayes (NB), Decision Tree (C5.0), Random Forest (RF). Small peptides sequences are selected from ARA-PEPs repository (Hazarika, et al., 2017) where 13748 small peptides are listed with six categories (i.e., secreted, sORF, stress-induced peptides (SIP), secreted-sORF, sORF-SIP, SIP-secreted). Total 27 features are fetched for each small peptides sequence to prepare data. Comparison is done using metrics i.e., F-Value, Sensitivity, Specificity, ROC, and FP rate with some statistical validation i.e., Kappa Statistics and Wilcoxon sign ranked test. Results of this study show that the proposed classifier has potential to accurately classify multi-level imbalanced data.

Nowadays, practically everyone, from a youngster to an adult, spends more time on online social media platforms, connecting with and exchanging information with individuals all over the world. The social network is becoming a popular means to communicate with people who live in different parts of the world. Because of the tremendous interconnectedness and information sharing enabled by the internet, social media platforms. This highlights the importance of establishing a system capable of detecting fake profiles on social media networks. There has been a lot of study done in this area utilising machine learning algorithms to identify fake profile, duplicate, spam, and bot accounts, and most of the fake profile accounts were effectively recognised using machine learning algorithms. This study discusses fake profile detection on social networks using machine learning.

Proteomic analysis of cancers' stages has provided new opportunities for the development of novel, highly sensitive diagnostic tools which helps early detection of cancer. This paper introduces a new feature ranking approach called FRMT. FRMT is based on the Technique for Order of Preference by Similarity to Ideal Solution method (TOPSIS) which select the most discriminative proteins from proteomics data for cancer staging. In this approach, outcomes of 10 feature selection techniques were combined by TOPSIS method, to select the final discriminative proteins from seven different proteomic databases of protein expression profiles. In the proposed workflow, feature selection methods and protein expressions have been considered as criteria and alternatives in TOPSIS, respectively. The proposed method is tested on seven various classifier models in a 10-fold cross validation procedure that repeated 30 times on the seven cancer datasets. The obtained results proved the higher stability and superior classification performance of method in comparison with other methods, and it is less sensitive to the applied classifier. Moreover, the final introduced proteins are informative and have the potential for application in the real medical practice.

In today's world, social media has spread widely, and the social life of people have become deeply associated with social media use. They use it to communicate with each other, share events and news, and even run businesses. The huge growth in social media and the massive number of users has lured attackers to distribute harmful content through fake accounts, leading to a large number of people falling victim to those accounts. In this work, we propose a mechanism for identifying fake accounts on the social media site Twitter by using two methods to preprocess data and extract the most effective features, they are the spearman correlation coefficient and the chi-square test. For classification, we used supervised machine learning algorithms based on the ensemble system (stack method) by using random forest, support vector machine, and naive Bayes algorithms in the first level of the stack, and the logistic regression algorithm as a meta classifier. The stack ensemble system was shown to be effective in achieving the best results when compared to the algorithms used with it, with data accuracy reaching 99%.

The DNA microarray technology has modernized the approach of biology research in such a way that scientists can now measure the expression levels of thousands of genes simultaneously in a single experiment. Gene expression profiles, which represent the state of a cell at a molecular level, have great potential as a medical diagnosis tool. But compared to the number of genes involved, available training data sets generally have a fairly small sample size for classification. These training data limitations constitute a challenge to certain classification methodologies. Feature selection techniques can be used to extract the marker genes which influence the classification accuracy effectively by eliminating the un wanted noisy and redundant genes This paper presents a review of feature selection techniques that have been employed in micro array data based cancer classification and also the predominant role of SVM for cancer classification.

Cancer is a world-famous deadly disease. According to the World Health Organization (WHO), cancer is the second leading cause of death globally and is responsible for an estimated 9.6 million deaths in 2018. One well-known technique for cancer detection is the DNA microarray technique. DNA microarray technology provides an opportunity for researchers to analyze thousands of gene expression profiles at the same time to determine whether a person has cancer or not. However, one of the problems in DNA microarray data is the large number of features that require feature selection. In overcoming these problems, this study will use the feature selection Support Vector Machine-Recursive Feature Elimination (SVM-RFE) and Chi-Square and use the Naive Bayes classification method. The accuracy results from using feature selection with those that are not will be compared. The accuracy between using the two feature selection methods will also be compared to find which feature selection method is...

Gene Selection from gene expression data for Cancer prediction has been an area of intensive research, aiming at identifying the minimal and optimal set of candidate genes that could generate accurate predictive performance. The two major problems encountered in this process are the high dimensionality of data with comparatively few instances and the need to categorize records under multiple classes. In this paper we propose a novel approach called Rank-Weight Feature Selection that utilizes the filtering capacity of more than one feature selection algorithm to detect the minimal set of predictive genes that generate higher predictor performance in categorizing and predicting diverse oncogenic gene expression data. The filtered features (genes) are weighted based on the number of feature relevance algorithms reporting them to be significant. The ranked genes are then used to validate the proposed method by utilizing ten classifiers over five diverse gene expression datasets. The results proved that the proposed approach generated higher predictive performance with fewer features than previously reported results with the most relevant and minimal set of genes and commend classifiers based on their accuracy and reliability in predicting cancer data.

Gene expression data is characterized by high dimensionality and small number of samples. Reducing the dimensionality is essential for effective analysis of the samples for efficient knowledge discovery. Actually, there is a tradeoff between feature selection and maintaining acceptable accuracy. The target is to find the reduction level or compact set of features which once used for knowledge discovery will lead to acceptable accuracy. Realizing the importance of dimensionality reduction for gene expression data, this paper presents novel framework which integrates dimensionality reduction with classification for gene expression data analysis. In other words, we present techniques for feature selection and demonstrate their effectiveness once coupled with data mining techniques for knowledge discovery. We concentrate on four feature selection techniques, namely chi-square, consistency subset, clustering-based and communitybased. The effectiveness of the feature reduction techniques ...

Microarray data classification is one of the most emerging clinical applications in the medical community. The classification process takes the detection of relevant and irrelevant probes into account, which is fundamental for subsequent classification. In this thesis, an efficient technique is proposed for the precise classification of microarray genes from the microarray gene expression dataset. The proposed classification technique performs the classification process with the aid of three phases namely, feature extraction, dimensionality reduction, and gene classification. Initially, Principal Component Analysis (PCA) is applied for dimension reduction and significant features are extracted from high dimension microarray data. The original data is projected to the lower dimension for selecting the eigenvector for the co-variance matrix shown cumulative variance up to a label of 100%. After the implementation of PCA, the reduced feature matrix so obtained is divided into two sets such as, training and test data set. Feed Forward Neural Network is a train using Back Propagation Algorithm and the performance accuracy of the classifier is tested using the test dataset (untrained samples). The total number of epochs required to reach the degree of accuracy is referred as the convergence rate. To check the efficiency of the algorithm the convergence rate is considered by plotting graphs between numbers of epochs vs. means square error (MSE). The two keys controlling parameters of BP algorithm are learning rate and momentum which affects the accuracy as well as the convergence rate. Varying the values of the parameters various graphs are plotted.

The microarray can track the expression levels of thousands of genes simultaneously. The high dimensional feature vectors of microarray impose a high dimensional cost as well as the risk of overfitting during classification. Thus it is necessary to reduce the dimension through ways like feature selection. Two basic approaches of feature selection appear: filter and wrapper techniques. In this paper, we make the people aware of the various techniques of feature selection. Keywords— Microarray, Feature Selection, Filter, Wrapper,

Data mining is the most popular knowledge extraction method for knowledge discovery. Machine learning is used to enable a program to analyze data, understand correlations and make use of insights to solve problems and enrich data and for prediction. Data mining techniques and machine learning algorithms play a very important role in medical area. The health care industry contains a huge amount of data. But most of it is not effectively used. Heart disease is one of the main reason for death of people in the world. Nearly 47% of all deaths are caused by heart diseases. We use KNN, Support Vector Machine to predict the heart diseases. Accuracy of the prediction level is high when using more number of attributes. Our aim is to perform predictive analysis using these data mining, machine learning algorithms on heart diseases and analyse the various mining, Machine Learning algorithms used and conclude which techniques are effective and efficient.

Classification of the different categories of small peptides is a challenging research area in bioinformatics research. However, machine learning based approaches are widely experimented in the literature with enormous success. For excellent learning of the classifiers, few numbers of informative features are important. This research explores a comparative study between various supervised feature selection methods such as Document Frequency (DF), Chi-Squared ( 2  ), Information Gain (IG), Gain Ratio (GR), Relief F (RF), and One R (OR). The corpus of small peptides data is selected from ARAPEP repository. Bayesian Classifier is taken to classify the different categories of the given corpus with the help of features selected by above feature selection techniques. Results of this study shows that RF is the excellent feature selection technique amongst other in terms of classification accuracy and false positive rate whereas DF and 2  were not so effective methods. Bayesian classifier has proven its worth in this study in terms of good performance accuracy and low false positives. Small peptides Identification, Machine Learning Classifiers, Pattern Recognition.

Classification of the different categories of small peptides is a challenging research area in bioinformatics research. However, machine learning based approaches are widely experimented in the literature with enormous success. For excellent learning of the classifiers, few numbers of informative features are important. This research explores a comparative study between various supervised feature selection methods such as Document Frequency (DF), Chi-Squared ( 2  ), Information Gain (IG), Gain Ratio (GR), Relief F (RF), and One R (OR). The corpus of small peptides data is selected from ARAPEP repository. Bayesian Classifier is taken to classify the different categories of the given corpus with the help of features selected by above feature selection techniques. Results of this study shows that RF is the excellent feature selection technique amongst other in terms of classification accuracy and false positive rate whereas DF and 2  were not so effective methods. Bayesian classifier has proven its worth in this study in terms of good performance accuracy and low false positives. Small peptides Identification, Machine Learning Classifiers, Pattern Recognition

Classification of the different categories of small peptides is a challenging research area in bioinformatics research. However, machine learning based approaches are widely experimented in the literature with enormous success. For excellent learning of the classifiers, few numbers of informative features are important. This research explores a comparative study between various supervised feature selection methods such as Document Frequency (DF), Chi-Squared ( 2  ), Information Gain (IG), Gain Ratio (GR), Relief F (RF), and One R (OR). The corpus of small peptides data is selected from ARAPEP repository. Bayesian Classifier is taken to classify the different categories of the given corpus with the help of features selected by above feature selection techniques. Results of this study shows that RF is the excellent feature selection technique amongst other in terms of classification accuracy and false positive rate whereas DF and 2  were not so effective methods. Bayesian classifier has proven its worth in this study in terms of good performance accuracy and low false positives. Small peptides Identification, Machine Learning Classifiers, Pattern Recognition

Classification of the different categories of small peptides is a challenging research area in bioinformatics research. However, machine learning based approaches are widely experimented in the literature with enormous success. For excellent learning of the classifiers, few numbers of informative features are important. This research explores a comparative study between various supervised feature selection methods such as Document Frequency (DF), Chi-Squared ( 2  ), Information Gain (IG), Gain Ratio (GR), Relief F (RF), and One R (OR). The corpus of small peptides data is selected from ARAPEP repository. Bayesian Classifier is taken to classify the different categories of the given corpus with the help of features selected by above feature selection techniques. Results of this study shows that RF is the excellent feature selection technique amongst other in terms of classification accuracy and false positive rate whereas DF and 2  were not so effective methods. Bayesian classifier has proven its worth in this study in terms of good performance accuracy and low false positives. Small peptides Identification, Machine Learning Classifiers, Pattern Recognition

Recognition of genetic expression becomes an important issue for research while diagnosing genetic diseases. Microarrays are considered as the representation for identifying gene behaviors that may help in detection process. Hence, it is used in analyzing samples that may be normal or affected, also in diagnosing various gene-based diseases. Various clustering and classification techniques were used to face the challenges in handling microarray. High dimensional data is one of the major issues caused while handling microarray. Also because of this issue, possibilities of redundant, irrelevant and noisy data may occur. To solve this problem feature selection process which optimally extracts the features is introduced in clustering in classification techniques. This survey observes some various techniques of classification, clustering of genes and feature selection methods such as supervised, unsupervised and semi-supervised methods. To determine the suitable semi-supervised algorithm that combines and analyze for detecting new or difficult mutated disease. This survey shows that how semi-supervised approach evolves and outperforms the existing algorithms.

This work focuses on the combining a feature selection technique based on genetic algorithm and support vector machines (SVM) of medical disease classification. SVMs have been used for various applications as a powerful tool for pattern classification. We use evolutionary computation which is a subfield of artificial intelligence or computational intelligence that involves combinatorial optimization problems. Evolutionary computation uses iterative progress, such as growth or development in a population. This population is then selected in a guided random search using parallel processing to achieve the desired end. Such processes are often inspired by biological mechanisms of evolution. The proposed genetic-SVM classifier is used to the best subset of features that can optimize the SVM classifier. It was concluded that the proposed genetic-SVM with a existing approach performs generally betterl and shown that our proposed method been very powerful for different data sets. The obtained results using the genetic algorithms approach show that the proposed method is able to find an appropriate feature subset and SVM classifier achieves better results than other methods. I. INTRODUCTION The classification of medical data has become an increasingly challenging problem, due to recent advances in medical mining technology. Classification of this tremendous amount of data is time consuming and utilizes excessive computational effort, which may not be appropriate for many applications. In this work, we develop an approach to optimize the support vector machine parameters which combines the merits of support vector machine (SVM) and genetic algorithm (GA). Later we compare our proposed method with the statistical approach to show its superiority in terms of computational efficiency. Feature selection is often an essential data pre-processing step prior to applying a classification models. Feature selection has been an active research area in pattern recognition, medical diagnosis, statistics, and data mining communities The objective of feature selection is to select a subset of relevant features for building effective prediction models. By removing irrelevant and redundant features, feature selection can improve the performance of prediction models by alleviating the effect of the curse of dimensionality, enhancing the generalization performance, speeding up the learning process, and improving the model interpretability. Support vector machines (SVMs) have been used for various applications as a powerful tool for pattern classification. The success of SVMs is based on (1) mapping the input space to a high-dimensional feature space, and (2) the maximization of the margin between two classes in the feature space. One of the advantages of SVMs is that we can improve generalization ability by proper selection of kernels. This paper is structured as follows. Section 2 introduces our methodology for classification, recent developments for feature selection, Genetic Algorithm and SVMs are reviewed. Several important aspects that arise from this work are discussed in Section 3. Section 4 provides experimental results using five real-world UCI data sets are discussed. A summary of this paper can be found in Section 5 where we provide its main conclusions and address future developments. II. METHODOLOGY A.Feature Selection Feature selection can be defined as a process that chooses a minimum subset of M features from the original set of N features, so that the feature space is optimally reduced according to a certain evaluation criterion. Feature selection is a fundamental problem in many different areas, especially in medical diagnosis, document classification, bioinformatics, and object recognition or in modeling of complex technological processes [3, 4, 12]. The main idea of feature selection is to choose a subset of input variables by eliminating features with little or no predictive information. Feature selection can significantly improve the comprehensibility of the resulting classifier models and often build a model that generalizes better to unseen points. Further, it is often the case that finding the correct subset of predictive features is an important problem in its own right. For example, physician may make a decision based on the selected features whether a dangerous surgery is necessary for treatment or not.