Abdulaziz Alhamadani, Shailik Sarkar, Lulwah AlKulaib


Drug overdose deaths are a dreadful crisis that leads to substantial societal impairments. Its harmful impact directly affects families and communities. To assist policymakers in mitigating this crisis, it is crucial to study the societal, economic, and criminal contributing factors linked to the crisis. Unfortunately, current data-driven works assume a singular factor, such as poverty being the cause and disregarding other realistic causes. Besides, recent works exhibited a lack of explainable models and spatial analysis of the crisis. Thus, DOD-Explainer links the gap by developing a realistic framework that predicts highly impacted counties of drug overdose deaths from crime and socioeconomic data. DOD-Explainer overcomes the challenge of data scarcity by proposing three data augmentation methods. Then, an algorithm is proposed to provide realistic explanations of the leading causes of the crisis. The results show that our application achieves the best predictive accuracy from several models, accurately identifies the most/least impacted counties by the crisis, and reveals the most contributing factors of drug overdoses.

Abdulaziz Alhamadani, Shailik Sarkar, Lulwah Alkulaib, Chang-Tien Lu: DOD-Explainer: Explainable Drug Overdose Deaths Predictor from Crime and Socioeconomic Data. IEEE Big Data 2022: 5163-5172


Publication Details

Date of publication:
January 26, 2023
Big Data
Page number(s):