Improved Jaya Optimization Algorithm for Feature Selection on Cancer Diagnosis Data using Evolutionary Binary Coded Approach

Abstract

Cancer Diagnostic Decision Support System may mislead the classification algorithms by
multiplicity of features. Hence, feature selection becomes an essential step in data mining of cancer
diagnosis data. Feature selection is a multi-objective optimization problem that systematically selects
a subset of most informative features for model building. Recently, randomization based
Evolutionary Algorithms (EAs) are becoming popular for feature selection than traditional filter and
wrapper methods. EAs explore the entire search space using some heuristic techniques in less time.
Jaya Optimization Algorithm (JOA) is a competitive Swarm Intelligence (SI) based EA, yet is simple
and easy to implement for solving optimization problems. The present work proposes BinJOA-S
Algorithm, an improved binary variant of JOA for solving feature selection problem on cancer
diagnosis data using Sigmoidal function. BinJOA-S algorithm employs the scalarization method to
solve feature selection as a multi-objective problem. The experimental results show that the
algorithm can obtain competitive performance when evaluated for parameters such as best, average
and worst fitness, average classification accuracy, best classification accuracy, the average number of
features selected and, CPU computational time. It is observed that the proposed algorithm produces
the effective performance of the cancer diagnosis system