Optimizing LSTM for Traffic Accident Prediction Using Beluga Optimization Algorithm

Abstract

Accurate prediction of traffic accidents plays a crucial role in enhancing road safety and mitigating risks. Long Short-Term Memory (LSTM) networks have demonstrated significant potential in modeling sequential data for traffic prediction tasks. Building on prior research that highlights the efficacy of Multi-LSTM models and convolutional neural networks, this study leverages the Beluga optimization algorithm to refine LSTM models for improved accuracy in predicting short-term traffic accidents. The proposed method addresses limitations related to data sources and factor analysis by enhancing parameter selection and model performance. Experimental results indicate that the optimized LSTM model surpasses conventional approaches, providing more precise predictions and contributing to advancements in traffic management and accident prevention