The Impact of Optimization Techniques on Cancer Classification with Gene Expression Data

Cancer classification using gene expression profiling is a fundamental component of cancer research and development, because it contributes to the identification of cancer subtypes as well as the identification of proper treatments for patients. The purpose of this paper is to assess the efficiency in classification of different optimization algorithms. PCA, GA, FI, SFS, and VT methods are applied for the feature selection on gene expression data. The selected features are then passed into a Gaussian Naive Bayes classifier to classify it into a particular class. In order to compare the efficiency of each presented optimization technique we calculate various evaluation metrices, and CPU time. To overcome the shortcomings of both the high accuracy methods, we proposed a new approach that integrates VT and GA. The outcomes of this work help to contribute to the research of the best combination of optimization tools for accurate classification of cancer, which will help further develop the concept of precision medicine and individualized therapy.