Introduction: The widespread misuse of opioids and cannabis is a notable global public health concern. The substantial public health concern due to the misuse of opioids and cannabis, individually and concurrently, is associated with vast societal implications. Identification of risk factors for developing misuse of these substances is of utmost importance. This study aims at developing a machine learning-based model to classify groups of opioid or cannabis dependents using family, microsocial, and medical history variables, and to identify the most significant variables associated with each group.Methods: This naturalistic observational non-interventional study enrolled adult patients diagnosed with opioid use disorder, cannabis use disorder, or a combination of both. Machine learning models, including Stacking, Logistic Regression, Gradient Boosting, k-Nearest Neighbors (kNN), Naive Bayes, Support Vector Machines (SVM), Random Forest, and Decision Tree, were used to classify patients and predict their risk factors based on various personal history variables.Results: The patient groups showed significant differences in their working fields, marital status before and after the formation of drug addiction, substance misuse in relatives, family type, parent-child relationships, and birth order. They also differed significantly in fleeing from home and personality types. Machine learning models provided high classification accuracy across all substance dependence groups, particularly for the cannabis group (>90% accuracy). Significant differences were found among the complex misuse group, where individuals faced severe psychosocial issues originating from the familial environment, such as a history of fleeing home, coming from a single-parent family, and dominant parent-child relationships.Discussion: The methods used in this study provided robust and reliable assessments of the models' predictive performances. The results pointed to significant differences in familial and developmental factors between the three dependence groups. The complex dependence group showed more severe psychosocial issues originating from the family environment. This group also revealed a specific sequence of life events and conditions predictive of complex dependence. These findings highlight the importance of interventions that address risk factors across various life stages and domains. Conclusion: Early identification of high-risk individuals and understanding the risk factors can inform the development of effective interventions at both individual and societal levels, ultimately aiming at mitigating dependence risks and improving overall well-being. Further research with longitudinal designs and diverse populations are needed to increase our understanding of trajectory of addiction formation in order to deliver effective interventions for individuals at risk.