Prediction of Adverse Drug Reactions Using Improved Feature Selection and Modified Fuzzy based Variability Ratio Tuning for SVM Classifier

C. Rathika and Dr.V. Umarani

In the discipline of drug discovery, one of the crucial issues is to predict the side effects of the drugs. Either protein targets or chemical structures are employed by earlier studies to envisage the symptoms and side effects for specific disease. Even though various machine learning techniques are been recommended for predicting the side effects, yet there exists room for further enhancement. But however predicting the side effects of drugs is foremost an important area. In this work, the drug information is utilised to portray the drugs from the chemical outlooks that comprises of associated proteins and chemical substructures. Firstly, feature selection is performed by using Improved Uniform Distribution based Bat Algorithm (IUDBA). This algorithm is used to eradicate redundant and irrelevant features. Distinguishing critical dimensions and lessening irrelevant dimensions may aid in identifying the sources of side effects. And finally, the side effect prediction is performed by Modified Fuzzy based variability ratio tuning for support vector machine (MFVR-SVM) classifier using an appropriate set of data features selected from feature selection. The simulation results demonstrate that the proposed Modified Fuzzy based variability ratio tuning for support vector machine (MFVR-SVM) classification technique provides best drug side effect prediction.

Volume 11 | 12-Special Issue

Pages: 421-434

DOI: 10.5373/JARDCS/V11SP12/20193238