Illicit Drug Abuse Detection Using Deep Reinforcement Learning with Imbalanced Data

Drug abuse is a global problem that affects human health and social security. Machine-learning-based drug abuse detection methods suffer from class-imbalance issues. A new framework for addressing the challenge of imbalanced data classification by deep reinforcement learning (DRL) was described. Improvements were made in three aspects: the Q network, reward function, and update strategy of the model to improve the illicit drugs abuse (IDA) detection accuracy and balance the detection accuracy of the majority and minority classes. An IDA haematological difference dataset containing 13 blood features was used to conduct ablation experiments on the model to verify the effectiveness of the improvement. Additionally, multiple methods optimised for imbalanced datasets, such as undersampling, oversampling, Synthetic Minority Oversampling Technique (SMOTE), and balanced random forest, were compared with our model. The results show that our model achieved an accuracy of 83.72$$\%$$ in the IDA detection task, which is slightly better than that of the other methods.