Icssdaf_predicting High-Risk Road Accident Zones in Malawi Using Ensemble Machine Learning and Smartphone Telematics Data

Road traffic accidents remain a major public health and transportation challenge, causing approximately 1.19 million deaths globally each year, with low- and middle-income countries experiencing a disproportionate share of the burden. In Malawi, the absence of data-driven road safety systems limits the ability to identify high-risk road segments and implement proactive accident prevention measures. This study proposes a machine learning-based framework for predicting high-risk road accident zones by integrating smartphone telematics data with historical crash records. An ensemble learning approach combining XGBoost and Random Forest algorithms was developed and deployed as a cloud-based microservice for road-risk estimation. The model achieved a coefficient of determination (R²) of 0.91 and a classification accuracy of 82.3% on the experimental dataset. The analysis identified six potential accident hotspot zones along major road corridors in Malawi. Furthermore, a cross-platform mobile application was developed to collect sensor data, visualize hotspot locations, and provide advance warnings to drivers when approaching hazardous road sections. Feature importance analysis revealed that speed-related variables accounted for approximately 68% of the model’s predictive capability, highlighting the significant influence of driving behavior on accident risk. The findings demonstrate the feasibility of integrating smartphone sensing technologies and ensemble machine learning models for road safety analytics in resource-constrained environments. The proposed framework provides a foundation for intelligent accident-risk monitoring and decision support systems aimed at enhancing road safety in Malawi and similar developing-country contexts.