A Machine Learning Technique to Detect Counterfeit Medicine Based on X-Ray Fluorescence Analyser

Since so many sub-standard and fake medicines are being openly sold, the counterfeit medicines have become widespread. The forgers succeeded in imitating the genuine medicines and make them look like genuine ones. This paper has proposed an approach that based on analysing the Tenormin ^R 50mg medicine by using non-destructive X-Ray Fluorescence Technique. This technique has been proposed over other heavy chemical analyzing methods to detect counterfeit Tenormin® due to its speed and reliability. There are 10 samples of Tenormin tablets from different manufactures were tested. All samples contained the active element Atenolol 50 mg and other inactive elements. Moreover two supervised machine learning techniques; RBF Support Vector Machine (RBF-SVM) and K-Nearest Neighbor (KNN) are employed. These two supervised machine learning algorithms were proposed as a step to design an automated approach in order to determine fake from genuine Tenormin without a need for trained chemists. The results revealed that X-Ray Fluorescence Technique has discriminated three elemental composition samples which differ from other 7 samples. The results also revealed the SVM proposed approach outperforms the KNN based approach with an overall accuracy of 93%.