ABSTRACT: The study of Gene Expression Profiling (GEP) of cells and tissue has become a major tool for discovery in medicine. In GEP, Gene Expression Microarray (GEM) data categorization is a difficult task because of its tremendous quantity of attributes (features) and also the limited size of sample. Feature Selection (FS) was explored to reduce the data dimensionality while maintaining the classifier accuracy. Recently, Swallow Swarm Optimization with Score-Based Criteria Fusion (Optimized SCF) wrapper FS approach has been suggested for predicting the tumors with better efficiency. However, relying upon simple statistical analyses or a unified FS might not increase prediction accuracy. Following the idea behind Ensemble FS (EFS), multiple algorithms are considered in this paper for increasing the robustness of classification. To give a contribution to the field, this Hybrid EFS (HEFS) system is introduced which encompasses different kinds of selection algorithms such as filter by Score-Based Criteria Fusion (SCF) and embedded method by Fuzzy Elephant Herding Optimization (FEHO) and Support Vector Machine- t (SVM-t). The outcomes from those approaches are aggregated via the Weighted Majority Voting (WMV). WMV is a popular and robust strategy to aggregate different algorithms, where each result according to their classification accuracy. The outcome of EFS is an attribute subgroup obtained by concatenating the results of different approaches on various data. It could improve the efficiency of the classifiers such as K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Recursive Neural Networks (RNN), and validate its superiority with four different datasets. Experimental results verify that the HEFS method shows improved results regarding precision, recall, accuracy and Area Under Curve (AUC) when compared to conventional FS methods.