Artificial Neural Networks in Drug Addiction Diagnosis

Abstract

This study aims to find a simple mechanism to help researchers and families identify addicts. In this paper, the Artificial Neural Network (ANN) method has been examined to determine whether a person is an addict. In this study, the dataset obtained from students from different countries and published as open source by Atif Masih was used. This dataset contains 50343 samples with 11 features. The study involved testing and comparing multiple neural network architectures based on their average classification accuracy. When the correlation matrix is examined, it is seen that the relationships between the variables are almost negligible. This can be attributed to the fact that the variables are categorical. Each architecture was trained using 10 different seed numbers, and the mean accuracy was calculated accordingly. The experiment results have obtained 75.53% classification accuracy for correct diagnosis in our system. Our model could significantly expedite the diagnosis and treatment of addiction, providing a valuable tool for families, physicians, and investigators. The paper proposes a Decision Support System (DSS) for diagnosing addiction, leveraging one of the most widely-used machine learning techniques: Artificial Neural Networks (ANN).