In this study we focus on developing a stock trend prediction model based on a modified version of support vector machine, named volume weighted support vector machine, along with a hybrid feature selection method named FSSFS method. In order to evaluate the prediction accuracy of this model we compare the VW-SVM classifier with plain support vector machine along with three commonly used feature selection methods including Information gain, Symmetrical uncertainty and correlation-based feature selection, via paired t-test. As the model input, we use several technical indicators and statistical measures, calculated for 10 stocks. The results show that the VW-SVM, combined with the hybrid feature selection method, significantly outperforms plain SVM model to the problem of stock trend prediction. In addition our experimental result show that VW-SVM combined with F-SSFS has the highest level of accuracies and generalization performance in comparison with the other three feature selection methods. With these results, we claim that VW-SVM combined with F-SSFS can serve as a promising addition to the existing stock trend prediction.

In some real-world applications, there is high-dimensional data which has led to many computational challenges. Feature selection is an effective technique for data dimensionality reduction, which simplifies the model and improves its performance by selecting the appropriate subset of features. In many of these applications, labeling of data is costly and time consuming, leaving little labeled data available and large amounts of unlabeled data available. In such applications, semi-supervised feature selection methods perform the feature selection process using the information of labeled data, and the distribution and geometric structure of labeled and unlabeled data. In most semi-supervised feature selection methods, a neighborhood graph is created and the importance of features is evaluated via their ability to maintain the geometric structure of the graph. In classical graph-based semi-supervised feature selection methods, the features are evaluated one by one and the correlation between features is not considered in feature selection process. To overcome this problem, sparse feature selection methods have been presented which consider the correlation between features, and calculate the optimal sparse transformation matrix for feature selection. In this paper, we investigate the semi-supervised learning methods, and review the graph-based semi- supervised sparse feature selection methods which select the appropriate features using the graph created by the labeled and unlabeled data, and the sparse regularization term. These methods solve the problem of classical semi-supervised methods by considering the correlation between features, create a neighborhood graph using the labeled and unlabeled data, calculate the graph Laplacian matrix, and compute the optimal sparse transformation matrix for feature selection.

- Ensemble feature selection methods are used to improve the robustness of feature selection algorithms. These approaches are a combination of several feature selection methods to achieve the final ranking of features. The reason for using such approaches is derived from the fact that the variety of different methods is more effective than only one method. Each feature selection algorithm may find feature subsets that can be considered local optima in the feature subsets space. Ensemble feature selection is a solution to address this problem. In this paper, we have proposed a bi-objective feature selection algorithm based on Pareto-based ranking. The maximum relevancy and minimum redundancy are considered as our two objectives. Both of the objectives are obtained by the ensemble of three feature selection methods. The final evaluation of features is according to a bi-objective optimization process and the crowding distance of features in this space for ranking the features. The proposed method results are compared with recent ensemble feature selection algorithms and simple feature selection algorithms. The results show that our classification accuracy method is superior to other similar methods and performs in a short time.

Feature selection is the process of selecting a subset of features among the set of primary features, so that, by removing the redundant and irrelevant features, the accuracy of the classification increases. Because of the low computational complexity, scalability in term of data dimensions and independence of any classifier, filter selection methods are very important. But one of the weaknesses of these methods is the lack of information about the interaction and communication between the features which leads to select redundant and irrelevant features. Selection of redundant and irrelevant features is due to the inappropriate selection of an objective function which estimates the significance and redundancy of the features. In this paper, a nonlinear filter feature selection method, based on conditional mutual information and Pareto set is presented and to prove the efficiency of it a series of experiments are performed on twelve widely used datasets. According to the results, the proposed method is more accurate than a number of recently feature selection methods.

Due to the increase in the volume of data and information in recent years, the issue of choosing the most appropriate feature for decision making has become very important. Classic attribute selection methods cannot work well on big data. Because feature selection is a complex problem, it seems appropriate to use meta-heuristic algorithms to solve this problem. In this paper, a new meta-heuristic algorithm inspired by nomadic migration to solve the feature selection problem is presented. This algorithm is named in honor of the Qashqai tribe. In this hybrid algorithm, the proportional function was designed based on the feature selection algorithm and based on minimizing the number of features and the amount of data error using neural network results. Then the Qashqai meta-heuristic algorithm was implemented on this fitness function and the results were compared with the well-known meta-heuristic algorithms of genetics and particle swarm. The results of the hypothesis test showed that the Qashqai optimization algorithm to solve the feature selection problem by the genetic algorithm and particle swarm is not defeated and in terms of convergence to the optimal solution works well.

Feature selection is one of the most important techniques in machine learning and pattern recognition, which eliminates redudant features and selects a suitable subset of features. This avoids overfitting when building the model and improves the model performance. In many applications, obtaining labeled data is costly and time consuming, while unlabeled data are readily available. Therefore, semi-supervised feature selection methods can be used to consider both labeled and unlabeled data in the feature selection process. In this paper, a semi-supervised sparse feature selection method is proposed based on hessian regularization and Fisher discriminant analysis which selects the appropriate features using the labeled data and the local structure of both labeled and unlabeled data. In the proposed method, an objective function based on semi-supervised scatter matrix and l2,1-norm is presented for feature selection which considers the correlation among features. To solve the proposed objective function, an iterative algorithm is used and its convergence is experimentally and theoretically proved. The results of the experiments on five data sets indicate that the proposed method improves the selection of relevant features compared to other methods used in this paper.

Feature selection methods are known to be effective in improving the learning process. The purpose of a feature selection method is to identify relevant features and remove irrelevant features in order to obtain a suitable subset of features, so that the redundancy between the selected features is minimized. In multi-label data, if there is a correlation between features, it is possible that the amount of redundancy in the feature set is increased. The existence of redundancy between features along with the challenge of high dimensions of multi-label data can grow the computational calculations, decrease the accuracy and finally increase the probability of errors in the prediction and classification of multi-label data. In this article, with the aim of minimizing the redundancy of features, a multi-label feature selection algorithm is proposed considering the least squares regression model and sparse regularization. Finally, using a number of well-known multi-label data sets, the efficiency of the proposed method is verified and the results are compared with some common multi-label feature selection methods.

Feature selection is one of the key challenges in developing intrusion detection systems. Classification algorithms in intrusion detection systems may be inconvenient for problems having so many features, because the size of the search space grows exponentially in terms of the number of features. This is while most of the features may be either irrelevant or redundant. Therefore, considering only relevant features (i.e. feature selection) may have a significant impact on the performance of the classification algorithms. The Imperialist Competitive Algorithm (ICA) can be used as a feature selection method with a high convergence, but it sometimes gets trapped in a local optimum. On the contrary, the Genetic Algorithm (GA) is powerful enough in terms of search for solutions, but it suffers from late convergence. Therefore, using a combination of both algorithms for feature selection may result in a rapid convergence as well as in a high precision. In this paper, by applying the Assimilate operator of the ICA to the GA, we propose a new feature selection algorithm for intrusion detection systems. The proposed algorithm has been tested on the KDD99 dataset using the decision tree classification. The experimental results show that the proposed algorithm has improved the detection rate (95.03%), false alarm rate (1.46) and the speed of convergence (3.82 second).

Financial distress prediction (FDP) is a great important subject that has always been interesting to researchers, financial institutions and banks. Tough many works have been done in this area, but use of combined approach of feature selection and classifier is an issue that has attracted researcher's attention just in recent years. In this paper, four well-known kinds of SVM that each of them has it's own kernel function including: linear, polynomial, radial and sigmoid have been introduced as the main classifiers of our proposed approach. These four methods have been integrated with genetic algorithm (GA) as a wrapper feature selection approach as well as three techniques of filtering feature selection approach called: principle component analysis (PCA), information gain and relief. Brought results indicated that genetic algorithm outperformed the other feature selection techniques in it's combination with SVM methods. Furthermore, implemented hypothesis test implied that there was no significance level among GA-SVM (linear), GA-SVM (radial), GA-SVM (polynomial) and GA-SVM (sigmoid) techniques with confidence level of %95.

Feature selection is a data preprocessing technique used for high-dimensional data sets before machine learning and data mining algorithms. The feature selection aims to find a minimal and optimal subset of the feature set. This subset includes valuable features while not including redundant ones. To do this, many current feature selection methods require the entire feature at first, and if a new feature is added to the feature set in the future, the algorithm must be run from the beginning. However, it is impossible to get all the features in many real-world applications or even wait for them. Therefore, online feature selection methods are provided for such issues that the entire feature space is not available at first. This paper presents an online feature selection method using the concept of Choquet fuzzy integral. This method first evaluates feature flows based on several filter criteria. Then, based on the Choquet operator, their results are combined, and decisions are made to preserve or ignore the feature. In the evaluation step, the performance of the proposed algorithm is compared with six online feature selection methods based on two categories. The proposed method is based on the results obtained in five real-world datasets that achieve about two percent improvement over similar methods based on classification accuracy and F-Score criteria. Also, due to the simple calculations in the process of the proposed method, the evaluation of features is done in a short time.