Traditional Feature Selection Algorithms Research Articles

An adaptive dual-strategy constrained optimization-based coevolutionary optimizer for high-dimensional feature selection

The feature subset obtained by traditional feature selection algorithms usually contains many irrelevant features and redundant features, which increases the size of the feature set and reduces classification accuracy. More importantly, these irrelevant features and redundant features may lead to local optimum. How to efficiently select the global optimal feature subset from high-dimensional data is a challenging task. To solve the problem, an adaptive dual-strategy constrained optimization-based coevolutionary optimizer for high-dimensional feature selection (ADSCO) is proposed in this paper. The contributions of this algorithm are as follows. Firstly, an adaptive dual strategy constrained optimization mechanism (ADSC) is developed. In the ADSC, the adaptive linear constraint strategy and the random constraint strategy are designed to search for the feature subsets that satisfy the constraints. This can reduce the search space of candidate solutions. Secondly, the dual population-constrained coevolutionary optimizer (CO) is developed to help the algorithm jump out of the local optimum and search for better optimal solutions to further improve the classification accuracy of ADSCO. Finally, the ablation experiments are executed to validate the effectiveness of the proposed ADSC and CO. The ADSCO algorithm is tested on 8 high-dimensional datasets and 8 low-dimensional datasets, and compared with seven typical population feature selection algorithms. Experimental results show that the proposed ADSCO algorithm is superior to the 7 comparison algorithms in terms of classification accuracy and feature subset size, demonstrating its strong competitiveness in tackling high-dimensional data feature selection.

Multi-strategy enhanced Marine Predators Algorithm with applications in engineering optimization and feature selection problems

This paper introduces a novel meta-heuristic optimization algorithm named Clustering Wavelet Opposition-based Marine Predators Algorithm (CWOMPA) to address some limitations present in the well-established Marine Predators Algorithm (MPA). CWOMPA incorporates three key strategies: a fuzzy clustering approach for escaping local optima, using wavelet basis function-based impact coefficient adjustment for elites to prevent premature convergence, and finally an adaptive opposition-based learning strategy for maintaining population diversity. Compared with some recent meta-heuristic algorithms, extensive evaluations conducted affirm that CWOMPA achieves the best Friedman rank, 4.30 and 1.95 respectively, on 23 benchmark functions and the CEC 2017 benchmark set. Not only does CWOMPA demonstrate significant effectiveness on six constrained problems from the CEC 2020 real-world benchmarks, but also when applied to 14 medical datasets, it gains superior Friedman ranks in terms of selected features, classification accuracy, F-Score, and objective function value, 2.5, 2.25, 2.96, 2.93 respectively, outperforming other common methods. Finally, CWOMPA exhibits the highest classification accuracy across all datasets and the best F-Score performance in nine datasets compared to traditional feature selection algorithms. The obtained results reveal that CWOMPA is a powerful and versatile optimization algorithm with significant potential in various real-world applications, including feature selection for medical datasets.

An ensemble maximal feature subset selection for smartphone based human activity recognition

Smartphone-based Human Activity Recognition (HAR) uses spatiotemporal time series data collected from a smartphone’s in-built accelerometer, gyroscope, and magnetometer sensor. Real-time HAR datasets such as UCI-HAR and UCI-HAPT extract time and frequency domain statistical features for activity recognition from the collected time series signals. Various Machine Learning techniques have been proposed in the perspective of feature selection through ensemble, traditional, hybrid, and meta-heuristic-based techniques to improve the recognition performance. Though the feature selection techniques have shown promising results, the computational time and selection of valid significant features for better classification performance still need to be included. Moreover, these techniques are highly parameter-dependent and critical to threshold values used for experimentation. This research proposes a novel method of parameter-free ensemble feature selection by combining traditional feature selection algorithms with the concept of association rule mining to obtain the Maximal Feature Subset (MFS), which is used for training and testing the classification model for HAR. Mathematical association rule mining is framed rather than the traditional one to select the associated frequent feature(s) from the feature set evolved by traditional techniques. Experimentation has been carried out on UCI-HAR and UCI-HAPT, and the proposed method has increased the classification efficiency by selecting significant features and reduces computation time significantly.

Dynamic multi-label feature selection algorithm based on label importance and label correlation

Multi-label distribution is a popular direction in current machine learning research and is relevant to many practical problems. In multi-label learning, samples are usually described by high-dimensional features, many of which are redundant or invalid. This paper proposes a multi-label static feature selection algorithm to solve the problems caused by high-dimensional features of multi-label learning samples. This algorithm is based on label importance and label relevance, and improves the neighborhood rough set model. One reason for using neighborhood rough sets is that feature selection using neighborhood rough sets does not require any prior knowledge of the feature space structure. Another reason is that it does not destroy the neighborhood and order structure of the data when processing multi-label data. The method of mutual information is used to achieve the extension from single labels to multiple labels in the multi-label neighborhood; through this method, the label importance and label relevance of multi-label data are connected. In addition, in the multi-label task scenario, features may be interdependent and interrelated, and features often arrive incrementally or can be extracted continuously; we call these flow features. Traditional static feature selection algorithms do not handle flow features well. Therefore, this paper proposes a dynamic feature selection algorithm for flow features, which is based on previous static feature selection algorithms. The proposed static and dynamic algorithms have been tested on a multi-label learning task set and the experimental results show the effectiveness of both algorithms.

A hybrid feature selection algorithm combining information gain and grouping particle swarm optimization for cancer diagnosis.

Cancer diagnosis based on machine learning has become a popular application direction. Support vector machine (SVM), as a classical machine learning algorithm, has been widely used in cancer diagnosis because of its advantages in high-dimensional and small sample data. However, due to the high-dimensional feature space and high feature redundancy of gene expression data, SVM faces the problem of poor classification effect when dealing with such data. Based on this, this paper proposes a hybrid feature selection algorithm combining information gain and grouping particle swarm optimization (IG-GPSO). The algorithm firstly calculates the information gain values of the features and ranks them in descending order according to the value. Then, ranked features are grouped according to the information index, so that the features in the group are close, and the features outside the group are sparse. Finally, grouped features are searched using grouping PSO and evaluated according to in-group and out-group. Experimental results show that the average accuracy (ACC) of the SVM on the feature subset selected by the IG-GPSO is 98.50%, which is significantly better than the traditional feature selection algorithm. Compared with KNN, the classification effect of the feature subset selected by the IG-GPSO is still optimal. In addition, the results of multiple comparison tests show that the feature selection effect of the IG-GPSO is significantly better than that of traditional feature selection algorithms. The feature subset selected by IG-GPSO not only has the best classification effect, but also has the least feature scale (FS). More importantly, the IG-GPSO significantly improves the ACC of SVM in cancer diagnostic.

Causal Feature Selection With Dual Correction.

Causal feature selection methods aim to identify a Markov boundary (MB) of a class variable, and almost all the existing causal feature selection algorithms use conditional independence (CI) tests to learn the MB. However, in real-world applications, due to data issues (e.g., noisy or small samples), CI tests can be unreliable; thus, causal feature selection algorithms relying on CI tests encounter two types of errors: false positives (i.e., selecting false MB features) and false negatives (i.e., discarding true MB features). Existing algorithms only tackle either false positives or false negatives, and they cannot deal with both types of errors at the same time, leading to unsatisfactory results. To address this issue, we propose a dual-correction-strategy-based MB learning (DCMB) algorithm to correct the two types of errors simultaneously. Specifically, DCMB selectively removes false positives from the MB features currently selected, while selectively retrieving false negatives from the features currently discarded. To automatically determine the optimal number of selected features for the selective removal and retrieval in the dual correction strategy, we design the simulated-annealing-based DCMB (SA-DCMB) algorithm. Using benchmark Bayesian network (BN) datasets, the experimental results demonstrate that DCMB achieves substantial improvements on the MB learning accuracy compared with the existing MB learning methods. Empirical studies in real-world datasets validate the effectiveness of SA-DCMB for classification against state-of-the-art causal and traditional feature selection algorithms.

Copula entropy-based golden jackal optimization algorithm for high-dimensional feature selection problems

Feature selection (FS) is a crucial process that aims to remove unnecessary features from datasets. It plays a role in data mining and machine learning (ML) by reducing the risk associated with high-dimensional datasets. FS is considered a challenging problem that is difficult to solve efficiently due to its combinatorial nature. As the size of the problem increases, the computation time also grows. Recently, researchers have focused on metaheuristic FS algorithms specifically designed for high-dimensional datasets. Therefore, this article proposes a powerful metaheuristic algorithm called Binary Enhanced Golden Jackal Optimization (BEGJO), which is an improved version of the recently published Golden Jackal Optimization (GJO) algorithm. The original GJO algorithm faces challenges when dealing with high-dimensional FS problems, as it tends to get trapped in local optima. To address this issue, various enhancement strategies are employed to improve the efficiency of GJO. The proposed BEGJO algorithm utilizes Copula Entropy (CE) to reduce the dimensionality of high-dimensional FS problems while maintaining high classification accuracy using the K-Nearest Neighbour (K-NN) classifier. Additionally, four enhancement strategies are incorporated to enhance the exploration and exploitation capabilities of the fundamental GJO algorithm. The BEGJO algorithm is transformed into its binary form using the sigmoid transfer function, aligning it with the nature of the FS problem. It is then tested on various high-dimensional benchmark datasets. The effectiveness of BEGJO is evaluated by comparing it with well-known algorithms in terms of classification accuracy, feature dimension, and processing time. BEGJO outperforms other algorithms in terms of classification accuracy and feature dimension and ranks up to fourth in terms of processing time. Furthermore, the advantageous use of CE is demonstrated by comparing the performance of the proposed algorithm with traditional FS algorithms. Statistical evaluations are conducted to further validate the effectiveness and superiority of the proposed algorithm. The results confirm that BEGJO is an effective solution for high-dimensional FS problems.

Dynamic mutual information-based feature selection for multi-label learning

In classification problems, feature selection is used to identify important input features to reduce the dimensionality of the input space while improving or maintaining classification performance. Traditional feature selection algorithms are designed to handle single-label learning, but classification problems have recently emerged in multi-label domain. In this study, we propose a novel feature selection algorithm for classifying multi-label data. This proposed method is based on dynamic mutual information, which can handle redundancy among features controlling the input space. We compare the proposed method with some existing problem transformation and algorithm adaptation methods applied to real multi-label datasets using the metrics of multi-label accuracy and hamming loss. The results show that the proposed method demonstrates more stable and better performance for nearly all multi-label datasets.

Deep learning-based classification model for GPR151 activator activity prediction

BackgroundGPR151 is a kind of protein belonging to G protein-coupled receptor family that is closely associated with a variety of physiological and pathological processes.The potential use of GPR151 as a therapeutic target for the management of metabolic disorders has been demonstrated in several studies, highlighting the demand to explore its activators further. Activity prediction serves as a vital preliminary step in drug discovery, which is both costly and time-consuming. Thus, the development of reliable activity classification model has become an essential way in the process of drug discovery, aiming to enhance the efficiency of virtual screening.ResultsWe propose a learning-based method based on feature extractor and deep neural network to predict the activity of GPR151 activators. We first introduce a new molecular feature extraction algorithm which utilizes the idea of bag-of-words model in natural language to densify the sparse fingerprint vector. Mol2vec method is also used to extract diverse features. Then, we construct three classical feature selection algorithms and three types of deep learning model to enhance the representational capacity of molecules and predict activity label by five different classifiers. We conduct experiments using our own dataset of GPR151 activators. The results demonstrate high classification accuracy and stability, with the optimal model Mol2vec-CNN significantly improving performance across multiple classifiers. The svm classifier achieves the best accuracy of 0.92 and F1 score of 0.76 which indicates promising applications for our method in the field of activity prediction.ConclusionThe results suggest that the experimental design of this study is appropriate and well-conceived. The deep learning-based feature extraction algorithm established in this study outperforms traditional feature selection algorithm for activity prediction. The model developed can be effectively utilized in the pre-screening stage of drug virtual screening.

Combinatorial online high‐order interactive feature selection based on dynamic graph convolution network

Traditional feature selection algorithms assume that the sample and feature space is known before learning, while most of the data is feature streams or data streams in reality. Currently, streaming feature selection algorithms can retain relevant features by removing redundant and irrelevant features based on the interaction between features, but they ignore the specific number of features that have interaction. Most of the existing studies only consider the case of interaction between two features, which is not quite in line with most realistic scenarios, i.e., the number of features with interaction is unknown. This paper concentrates on the high-order interactions between stream feature, and proposes a Combinatorial High-Order Interactive Feature Selection based on Dynamic GCN and Sparse learning (CHOIFS-DGS). Based on previous definitions of feature interaction, this paper proposes some new metrics to measure the degree of interaction between a newly arrived feature and an already selected feature. CHOIFS-DGS consists of three main parts, namely: low-order online feature selection based on interaction measure, high-order online feature selection based on dynamic GCN, and Intra-group sparse feature selection. In the experimental analysis section, this study employs two different classifiers and eleven publicly released data sets, including gene data related to diseases and data from two classification challenges (NeurIPS 2003 feature selection challenge and WCCI 2006 Performance Prediction Challenge). The experimental results demonstrate that the proposed CHOIFS-DGS model significantly improves classification accuracy on all eleven data sets, while using a relatively smaller number of features, thus fulfilling the role of key feature selection. Furthermore, the CHOIFS-DGS algorithm consists of three components: LO-OIFS, HO-OIFS, and Group-Sparse. By applying these three sub-modules separately for extracting data features and comparing the results with CHOIFS-DGS, it is found that the performance of CHOIFS-DGS is lower than that of the individual sub-modules only in three data sets, while significantly better in the remaining eight data sets. This indicates that the integrated use of the three sub-modules can enhance model accuracy. Finally, in the ablation experiment, to verify the necessity of considering higher-order interactions among features, the results of the HO-OIFS module were compared with those of the other modules. The results show that the model’s accuracy significantly improves after incorporating the HO-OIFS module, thereby demonstrating that considering higher-order interactions between features is essential.

Adaptive graph regularization and self-expression for noise-aware feature selection

Many traditional unsupervised feature selection algorithms utilize manifold information to mine the local structure of the data. However, the noise existing in the raw data reduces the accuracy of the manifold information of data, which affects the learning effect of the entire algorithm. In order to solve the above problems and more fully find the internal structure inside the data, this paper proposes an adaptive graph regularization and self-expression for noise-aware feature selection (ASNFS). Firstly, the algorithm adopts non-negative matrix factorization to decompose the raw data matrix, and utilizes the low-dimensional matrix generated after decomposition to replace the raw high-dimensional data matrix. This allows the algorithm to reveal some internal structural information of the raw data while reducing the dimensionality of the data. ASNFS also introduces the orthogonal basis clustering with excellent clustering effect, and the interpretability of the algorithm is enhanced. Secondly, in addition to preserving the manifold information in the low-dimensional projection subspace, the algorithm also preserves the manifold information in the non-negative matrix factorization subspace. Meanwhile, the adaptive graph regularization term added to the objective function enables the algorithm to continuously update the similarity matrix. It can effectively remove the noise inside the raw data, and prevent the occurrence of over-fitting phenomenon of the experimental results caused by the fixed similarity matrix. Finally, the similarity matrix retains the local structure information of the data with each iteration, and the results of the feature selection are reused for the construction of the similarity matrix. ASNFS adopts an alternate iterative method to optimize the objective function, which is simple and effective. Then, the algorithm complexity and convergence are analyzed. ASNFS is compared with seven feature selection algorithms on nine datasets, and the experimental results reflect the effectiveness of ASNFS in feature selection.

Feature selection using Information Gain and decision information in neighborhood decision system

Feature selection is a significant preprocessing technique for data mining, which can promote the accuracy of data classification and shrink feature space by eliminating redundant features. Since traditional feature selection algorithms have high time complexity and low classification accuracy, an effective algorithm using Information Gain and decision information is designed. The algorithm introduces Information Gain for performing preliminary dimensionality reduction on high dimensional datasets, and then the decision information is regarded as an evaluation function of features to select features with important information. First, the concept of joint information granule is defined, and neighborhood information entropy measures are proposed based on the joint information granule. In addition, the relationship between these measures is studied, which is helpful to study the uncertainty in data. Second, a nonmonotonic algorithm using the decision information in the neighborhood information entropy measures is proposed to overcome the shortcoming of algorithms based on monotonic evaluation functions, thereby improving the accuracy of data classification. Third, to reduce the time cost of the designed algorithm for high dimensional datasets, Information Gain is introduced to preliminarily eliminate irrelevant features in high dimensional datasets. Finally, the ablation and comparison experiments on twelve public datasets demonstrate the low time cost and high classification accuracy of our algorithm, respectively.

DroidRL: Feature selection for android malware detection with reinforcement learning

Due to the completely open-source nature of Android, the exploitable vulnerability of malware attacks is increasing. Machine learning, leading to a great evolution in Android malware detection in recent years, is typically applied in the classification phase. Since the correlation between features is ignored in some traditional ranking-based feature selection algorithms, applying wrapper-based feature selection models is a topic worth investigating. Though considering the correlation between features, wrapper-based approaches are time-consuming for exploring all possible valid feature subsets when processing a large number of Android features. To reduce the computational expense of wrapper-based feature selection, a framework named DroidRL is proposed. The framework deploys DDQN algorithm to obtain a subset of features which can be used for effective malware classification. To select a valid subset of features over a larger range, the exploration-exploitation policy is applied in the model training phase. The recurrent neural network (RNN) is used as the decision network of DDQN to give the framework the ability to sequentially select features. Word embedding is applied for feature representation to enhance the framework’s ability to find the semantic relevance of features. The framework’s feature selection exhibits high performance without any human intervention and can be ported to other feature selection tasks with minor changes. The experiment results show a significant effect when using the Random Forest as DroidRL’s classifier, which reaches 95.6% accuracy with only 24 features selected.

GA-NobRFE-SVM: A New Algorithm for Classification of Unbalanced Gene Data

As an important branch of bioinformatics, gene microarray data analysis has become one of the important frontier fields in life sciences. Because of the high cost of microarray experiment, gene expression profile data shows the characteristics of small samples size, high dimensionality and category imbalance between samples. In this case, the traditional feature selection method is difficult to obtain good results. this paper proposes unbiased SVM-RFE (NobSVM-RFE). Compared with traditional feature selection algorithm, NobSVM-RFE algorithm can obtain better feature subset and reduce computation cost. Combining GASMOTE with NobSVM-RFE, this paper proposes a three-stage algorithm GA-NobRFE-SVM, which includes balancing algorithm, feature selection algorithm and classifier. The experimental results show that GA-NobRFE-SVM can effectively improve the classification performance of unbalanced gene data.

Bearing fault feature selection method based on dynamic time warped related searches

Traditional feature selection algorithms rarely consider the dynamic misalignment between different time series, and have poor fault tolerance and robustness. In this paper, a fault feature selection method for rolling bearings based on Dynamic Time Warped Related Searches (DTWRS) is proposed. Firstly, the bearing fault feature set is constructed, and the dynamic time warping algorithm is used to calculate the shortest cumulative distance between feature of different faults, and this distance is used as the correlation evaluation standard. Then, two new search rules, dynamic time warping difference and dynamic time warping entropy, are proposed based on the minimum redundancy between bearing fault features and the maximum correlation between fault features and feature categories, use these two search rules to judge the ability of the feature to express the fault, define the quality of the fault feature and sort from good to bad according to the level of ability. Finally, in this order, the number of features is gradually increased and input to the fault classifier, and the sensitive fault feature set is obtained based on the principle of the highest recognition rate and the least number of features. The experimental results show that the fault feature selection method of rolling bearing based on DTWRS can increase the accuracy of fault diagnosis while minimizing the number of features, and improve the efficiency and effect of fault diagnosis.

Research on Mutual Information Feature Selection Algorithm Based on Genetic Algorithm

<p>Feature selection is an important part of data preprocessing. Feature selection algorithms that use mutual information as evaluation can effectively handle different types of data, so it has been widely used. However, the potential relationship between relevance and redundancy in the evaluation criteria is often ignored, so that effective feature subsets cannot be selected. Optimize the evaluation criteria of the mutual information feature selection algorithm and propose a mutual information feature selection algorithm based on dynamic penalty factors (Dynamic Penalty Factor Mutual Information Feature Selection Algorithm, DPMFS). The penalty factor is dynamically calculated with different selected features, so as to achieve a relative balance between relevance and redundancy, and effectively play the synergy between relevance and redundancy, and select a suitable feature subset. Experimental results verify that the DPMFS algorithm can effectively improve the classification accuracy of the feature selection algorithm. Compared with the traditional chi-square, MIM and MIFS feature selection algorithms, the average classification accuracy of the random forest classifier for the six standard datasets is increased by 3.73%, 3.51% and 2.44%, respectively.</p> <p> </p>

Detection of soluble solid content in apples based on hyperspectral technology combined with deep learning algorithm

Soluble solid content (SSC) is the most important attribute related to the quality and price of apples. The objective of this study was to detect the SSC of “Fuji” apple using hyperspectral imaging (HSI) combined with a deep learning algorithm. A novel, output-related deep learning algorithm, stacked weighted auto-encoders (SWAE) was proposed to extract output-related features based on pixel-level spectra of each sample. Relevant variables were identified and assigned to different weights by correlation analysis with the output variable. To quantitatively predict the SSC under the constraint of preferential reconstruction for quality-related variables, the deep features containing information about the SSC quality prediction were extracted, and then these features were input into gray wolf optimization-support vector regression (GWO-SVR) to quantitatively predict SSC. Furthermore, successive projection algorithm (SPA) and competitive adaptive reweighed sampling (CARS) were adopted as traditional feature selection algorithms to build GWO-SVR models for predicting the SSC in apples and comparing them with the deep learning models. The results showed that the proposed SWAE-GWO-SVR model yielded satisfactory performance with R p 2 of 0.9436, and RMSEP of 0.1328 °Brix, which demonstrate that deep learning combined with HSI can facilitate the nondestructive assessment of the internal quality attributes of fruits. Practical applications Traditional methods for detecting the soluble solid content (SSC) in apples are destructive and laborious. Therefore, the hyperspectral imaging (HSI) technique combined with deep learning was used to determine the SSC in apples in a rapid and nondestructive manner. The results indicated that combined method is feasible for SSC prediction. Thus, HSI combined with deep learning is considered as a promising method for detecting the SSC in apples.

Multi-Cluster Feature Selection Based on Isometric Mapping

This letter presents an unsupervised feature selection method based on machine learning. Feature selection is an important component of artificial intelligence, machine learning, which can effectively solve the curse of dimensionality problem. Since most of the labeled data is expensive to obtain, this paper focuses on the unsupervised feature selection method. The distance metric of traditional unsupervised feature selection algorithms is usually based on Euclidean distance, and it is maybe unreasonable to map high-dimensional data into low-dimensional space by using Euclidean distance. Inspired by this, this paper combines manifold learning to improve the multi-cluster unsupervised feature selection algorithm. By using geodesic distance, we propose a multi-cluster feature selection based on isometric mapping (MCFS-I) algorithm to perform unsupervised feature selection adaptively for multiple clusters. Experimental results show that the proposed method consistently improves the clustering performance compared to the existing competing methods.

Sparse and low-dimensional representation with maximum entropy adaptive graph for feature selection

Traditional feature selection algorithms usually explore the relationship between data and cluster structure in a single space, so the internal relationship obtained is not very rich, and it is not enough to select more valuable features. To solve the above problems, in order to fully mine the intrinsic correlation information in different spaces, so that the relationship between data, features and clustering structure can be explored at the same time, this paper proposes a feature selection method based on sparse and low-dimensional representation with maximum entropy adaptive graph (SLMEA). Firstly, the SLMEA combines the sparse transform representation with pseudo-label matrix learning to optimize, and uses the pseudo-label matrix to guide the learning of sparse low-dimensional space. It can not only explore the relationship between data and pseudo-labels in data space, but also mine the association between features and pseudo-labels in feature space, so as to select more discriminative features. Secondly, based on the maximum entropy theory, the similarity matrix is constructed adaptively, so that the two different manifold structures corresponding to sparse transform representation and pseudo label matrix learning can be adaptively learned and retained in the iterative process. In addition, in order to ensure the sparsity of the transformation matrix, the constraint of ℓ2,1/2-norm is applied to the matrix, which can better deal with the redundant features and obtain more sparse solutions. Finally, the SLMEA uses an alternate iterative update method to optimize the objective function, and carries out extensive experiments on eight mainstream datasets. Compared with seven state-of-the-art algorithms, SLMEA can have higher clustering accuracy and normalized mutual information.

Informative Gene Selection Based on Cost-Sensitive Fast Correlation- Based Filter Feature Selection

Background: Informative gene selection is an essential step in performing tumor classification. However, it is difficult to select informative genes related to tumors from large-scale gene expression profiles because of their characteristics, such as high dimensionality, relatively small samples, and class imbalance, and some genes are superfluous and irrelevant. Objective: Many researchers analyze and process gene expression data to obtain classified gene subsets by using machine learning methods. However, the gene expression profiles of tumors often have massive computational challenges. In addition, when improving feature importance and classification accuracy, cost estimation is often ignored in traditional feature selection algorithms, which makes tumor classification more difficult. Methods: In this study, a novel informative gene selection method based on cost-sensitive fast correlation- based filter (CS-FCBF) feature selection is proposed. Results: First, the symmetric uncertainty index is used to evaluate the correlation between informativeFirst, the symmetric uncertainty index is used to evaluate the correlation between informative genes and class labels, then a large number of irrelevant and redundant genes are quickly filtered according to importance. Thereby, a candidate gene subset is generated. Second, costsensitive learning, which introduces the misclassification cost matrix and support vector machine attribute evaluation, is used to obtain the top-ranked gene subset with minimum misclassification loss. Finally, the candidate gene subset is optimized. genes and class labels, then a large number of irrelevant and redundant genes are quickly filtered according to importance. Thereby, a candidate gene subset is generated. Second, costsensitive learning, which introduces the misclassification cost matrix and support vector machine attribute evaluation, is used to obtain the top-ranked gene subset with minimum misclassification loss. Finally, the candidate gene subset is optimized. Conclusion: This experiment was verified in eight independent tumor datasets. By comparing and analyzing CS-FCBF with another three hybrids of typical gene selection algorithms combined with cost-sensitive learning, we found that the method proposed in this study has a better classification performance with fewer selected genes, which might provide guidance in tumor diagnosis and research.