Widespread of illicit Amphetamine-type Stimulants (ATS) drug abuse and trafficking has successfully gained international concern. With the growing issue of manufacturing new brand and unfamiliar ATS drug, a reliable method to identify and detect ATS drug is essential. Traditionally, ATS drugs analysis is carry out using laboratory testing, which is a costly and lengthy process. In this paper, we meant to propose an analytical method by adopting Artificial Intelligence approach. The main concern is to sort out the distinctive signature features that have high discriminative power in identifying ATS drugs. Therefore, this paper we address the problem of large feature subset from broad pattern of ATS and non-ATS drug data, which is extracted from its 3D molecular structure. Using the training samples, we develop an ensemble based of filter and embedded method that combines the output of Symmetrical Uncertainty (SU) method and Support Vector Machine - Recursive Feature Elimination (SVM-RFE) method. To achieve an optimum selection of feature set, we further apply a low variance filter on the ensemble feature set. The experimental evaluation will be perform using random forest classifier. The findings show that our proposed method can effectively reduce the number of features, and the classification accuracy can be enhance.