Temporal Pattern Mining in Illicit Drug Seizures: A Comparative Analysis of Heroin and Methamphetamine

The increasing number and changing trends of heroin and methamphetamine in Sri Lanka are a major problem for the health, law enforcement, and policy-making of the community. These illicit substances have been displaying significant growth and changes through the weight categories over the period between 2015 and 2024, indicating the necessity to intervene and solve the problem in a timely and data-driven manner. The analysis of data on seizures in this study was based on anomaly detection, change-point analysis, modeling of weight-category transition, and yearly clustering to reveal structural changes, trends, and patterns of interaction in drug-related activities. The cases of heroin increased steadily between the year 2015, when the number of cases per month was around 120 cases, and 2024, when the number of cases exceeded 450 cases per month, and the number of cases of methamphetamine increased after 2022 and had over 300 cases per month in 2024. The anomaly identification in June-September 2020 and January 2024 showed that the deviations were substantial, and this meant that the heroin activity was largely interrupted. The change-point analysis revealed the changes in the trends of heroin in 2018-2019 and 2025, but not the changes in methamphetamine, which began to change in 2022. Heroin incident at transitions between weight categories decreased (< 2 g) and increased (2-5 g and 5-10 g), indicating a trend of increasing typical seizure sizes. The annual clustering indicated that 2020 and 2022-2024 will be years of high activity of heroin, and 2024 will be a year of high activity of methamphetamine. This study will help offer a scalable decision-support model to policymakers, forensic labs, and law enforcement agencies by offering an integrated approach to drug activity change monitoring, strategic planning, and resource allocation. The framework also creates a base of predictive systems in the future that will take into account socio-economic, demographic, and real-time surveillance information to improve proactive intervention measures.