Artificial intelligence model for the prediction of cannabis addiction

Abstract

A novel approach for predicting cannabis addiction has been introduced by integrating combined machine learning (ML) algorithms, specifically
K-means clustering and linear regression (LR). The study, conducted in Marrakech, Morocco, at a center linked to the National Association for drug-risk reduction (DRR), involved 146 participants. Among those with prior cannabis use, one subgroup included passive users, while another exhibited cannabis dependence. The research utilized features derived from patient data, emphasizing psycho-cognitive state, addiction status, and socio-demographic factors. The goal was to evaluate the effectiveness of the combined ML algorithms (K-means + LR) in distinguishing between addicted and non-addicted individuals using real-world data from a primary care addiction center. The findings indicate that the proposed method delivers balanced results, achieving an overall accuracy of 70%, a sensitivity of 65%, and a specificity of 86%. These results are particularly noteworthy when compared to other ML studies in addiction research. The combined algorithm demonstrates promising potential with competitive accuracy and high specificity. Further efforts to improve sensitivity and validate the model in diverse settings will be essential for advancing predictive modeling in this field. Our findings contribute to existing research by developing simple and effective tools for early detection of cannabis addiction, potentially aiding in the creation of preventive and therapeutic strategies to reduce its prevalence.