Abstract

Machine learning (ML) holds promises in predicting potential substance abuse risk, enabling early interventions to prevent addictive behaviors. Most studies focus on prognosis of individuals already affected by substance use disorders (SUDs), however, a few investigations explore the prediction of future SUD risk. These studies use diverse data, including socio-economic status, psychological features, genetic information and social media activity, achieving accuracy rates more than 96% in some cases. However, heterogeneity in methodology and lack of standardized frameworks limit their applicability. This review highlights the need for consistent approaches to fully know the potential of ML in preventing substance abuse.