Spatial Prediction And Safe Route Optimization of Autonomous Vehicle Accidents: Machine Learning And Gis-Based Integrated Approach

Abstract

Autonomous vehicle technologies are reshaping traffic safety and urban mobility; however, software and sensor failures, legal uncertainties, and mixed traffic environments continue to pose significant accident risks. Recent fatal incidents during real-world testing underline the critical need for effective risk prediction approaches. While existing studies mainly focus on accident causes and probabilities, deep learning research is often limited to in-vehicle sensor data.This study proposes a hybrid framework integrating machine learning and deep learning models with Geographic Information Systems (GIS)-based spatial analysis to predict both the occurrence and severity of autonomous vehicle accidents and to identify high-risk locations and conditions. The results show that Support Vector Machine (SVM) and Artificial Neural Network (ANN) models perform effectively in accident severity prediction, while ANN achieves superior performance in accident occurrence prediction under balanced data conditions. The proposed approach offers a data-driven basis for risk-aware route planning and dynamic safety management in intelligent transportation systems.