Optimal Feature Selection from High-dimensional Microarray Dataset Employing Hybrid IG-Jaya Model

Abstract

This paper proposed a hybrid information gain and a Jaya algorithm-based model to identify the informative genes from the high dimensional microarray data set. Metaheuristic algorithms need to tune the parameters to achieve better accuracy, and it is a tidy and sensitive job for all researchers. To solve the difficulties mentioned above, we proposed IG-Jaya, a new hybrid FS model based on wrapping information gain with the Jaya optimization algorithm (parameterless) to obtain the optimal features from the microarray data set. The objective behind considering Jaya is to minimize the computing cost and the risk of tuning the algorithm’s parameters to achieve better accuracy. This algorithm’s main strength is that it identifies the best feature subset by updating the worst ones. The resulting feature subset can be considered an input for the classification model.The primary contributions of the study are 1.Using IG as a technique, we developed a filter-based paradigm to employ feature selection by removing redundant and irrelevant features from microarray cancer datasets. 2. This new hybrid meta-heuristic FS model, namely IG-Jaya, is proposed for the efficient diagnosis of cancer disease. 3. Different metrics such as sensitivity, specificity, accuracy, and AUC-ROC Curve are used to study the performance of the hybrid model with various classifiers such as SVM, LDA, DT, NB, etc. This paper’s overall investigation is divided into two phases: In the first part, without including any filter, we have used the parameter-less JA to identify featured gene subsets. And the performance of JA is evaluated using various classifiers like SVM, LDA, NB, and DT. From the resulting study, it is noteworthy to state that IG-JAYA performs better as compared to the existing models This paper proposed a hybrid information gain and a Jaya algorithm-based model to identify the informative genes from the high dimensional microarray data set. The performance evaluation of the proposed model is done with 13 different benchmark data sets. To achieve better performance, we have focused on one of the best meta-heuristic parameter-less algorithms called JAYA. It used the solution''s fitness to gather the most feasible informative genes. And from the comparison table, we can also ensure the model''s performance. For some datasets, our proposed model cannot provide the best accuracy compared to other existing approaches; it is pretty steady and sound. The same model will be tested with different filter methods and real-time datasets in the subsequent study. A hybrid multi-filter Jaya algorithm will be proposed to check the efficiency of the proposed one. And it would be better to choose any other hybrid model with JAYA to enhance the feature selection accuracy with a high dimensional dataset. In the future , it would be better to choose any other hybrid model (chaos-based) with JAYA to enhance the feature selection accuracy with a high-dimensional dataset.