The widespread problem of drug addiction remains a major public health issue in Bangladesh, demanding a thorough exploration of its root causes and predictive models. Data was gathered from survey responses of individuals online, covering a diverse set of demographic and behavioral characteristics. To evaluate their predictive potential, the data was examined using machine learning methods such as Linear Regression, Logistic Regression, Decision Tree, K-Nearest Neighbor Classifier, and Naïve Bayes Classifier. Our results demonstrate that the machine learning models serve as effective tools for comprehending and forecasting drug addiction within the Bangladeshi context. The performance metrics for each algorithm are as follows: Linear Regression (with a Mean Absolute Error of 0.2279 and Mean Squared Error of 0.0947), Logistic Regression (86 % accuracy), Decision Tree (86 % accuracy), K-Nearest Neighbor (KNN) (79% accuracy), and Naïve Bayes (86 % accuracy). These models exhibit promising capabilities, with Logistic Regression, Decision Tree, and Naïve Bayes notably achieving an 86 % accuracy rate in predicting drug addiction. These findings establish a solid foundation for future investigations into the complexities of drug addiction and the creation of more sophisticated predictive models specifically tailored to the context of Bangladesh.