Analysis of Feature Selection Method for 3D Molecular Structure of Amphetamine-Type Stimulants (ATS) Drugs

Abstract

AbstractThis paper proposes an improved method to analyze the effectiveness of ATS drugs identification by using a few feature selection methods such as Sequential Forward Floating Selection (SFFS), Sequential Forward Selection (SFS), Sequential Backward Floating Selection (SBFS), Sequential Backward Selection (SBS) and Support Vector Machine-Recursive Feature Elimination (SVM-RFE). The fundamental target of this paper is to compare which feature selection methods have better classification accuracy performance in identification for a large dataset. A comprehensive verification using WEKA is carried out to determine the performance of classification accuracy. This is achieved by comparing several classifiers with all features (without feature selection methods) and with selected features (with feature selection methods). From the experimental work, it was found that the performance of classification accuracy with selected features has similar accuracy if the performance accuracy done with all features. This shows that feature selection methods help to fasten and get better accuracy performance. The result also indicates that SFFS are the best feature selection methods to use to embed with SVM-RFE, while J48, IBk and Random Forest (RF) are the best three classifiers to use for future evaluation.KeywordsAmphetamine-Type Stimulants (ATS)As Sequential Forward Floating Selection (SFFS)3D molecular structureDrug image recognitionFeature selection method