Hybrid Brain Storm Optimization based Feature Selection and Optimal Deep Neural Network for Medical Data Classification

B. Prabadevi and N. Balakumar

Presently, medicinal data offers particular features which finds helpful to perform various processes like diagnosis and medication. Various medicinal classification problems are available and the choice of features present in the medical dataset remains a significant issue. The feature selection (FS) process is highly essential in the data classification problem, which helps to select the needed features from the available data results in less computation time and high classification accuracy. This paper devises a novel hybrid brain storm optimization (HBSO) based FS and optimal deep neural network (ODNN) based classification model for medicinal data, abbreviated as HBSOODNN. The brain storm optimization (BSO) algorithm is tuned by genetic algorithm (GA) to reduce the computation time and avoid local optima problem. The HBSO-GA model is applied for FS and the feature reduced subset is provided for data classification. To classify the data, the DNN fine tuned by particle swarm optimization (PSO) algorithm takes place, called as PSO-DNN model. The presented HBSO-ODNN model is validated using a set of four medicinal dataset and the results are validated under several aspects. The obtained simulation outcome stated that the presented model shows excellent classification outcome over the compared methods.

Volume 12 | 04-Special Issue

Pages: 62-72

DOI: 10.5373/JARDCS/V12SP4/20201467