Ensemble Filter-Embedded Feature Ranking Technique (FEFR) for 3D ATS Drug Molecular Structure

Abstract

The concern for illicit abused and trafficking of ATS drugs are continuously growing. This is due to the evolving of new and unfamiliar ATS drugs, present a significant challenge to the forensic staff and laboratory testing. This paper aims to explore the use of machine learning method in the 3D molecular structure of ATS drug identification. In order to perform the computational analysis, the 3D molecular structure of ATS drugs will be illustrated in the voxel format of data representation. This paper proposes a new ensemble feature selection technique of Filter-Embedded Feature Ranking Techniques (FEFR), which is the combination of the filter method (ReliefF) and embedded methods (Variable Importance based Random Forest). It is used to identify a subset of significant features with highly discriminative power in representing the molecular structure of ATS drugs. These selected significant features eventually improve the performance of identification task.