A GAN Based Approach for Predicting Traffic Accident Severity and Hospital Recommendations

Highway traffic accidents remain a major cause of death, even with improvements in traffic safety. In developing countries, the number of lives lost and the property damage from road accidents is particularly high. This study aims to create a GAN system that accurately predicts the severity of injuries from road accidents and recommends the most suitable hospital for treatment based on the type of injury. Traditionally, injury assessment and hospital recommendations depended on first responders or emergency personnel, which could delay urgent care. These systems lacked the speed and accuracy needed for timely injury evaluation, sometimes leading to delayed or incorrect treatment. The problem lies in the inability to quickly and accurately assess injuries without machine learning models, which can result in preventable deaths due to treatment delays. The proposed system uses GAN (Generative Adversarial Network) Techniques like WGAN (Wasserstein Generative Adversarial Networks), CGAN (Conditional Generative Adversarial Network), AC-GAN (Auxiliary Classifier GANs) can help in predicting factors related to crash severity to classify injuries (such as head, hand, or leg) and assess their severity based on the injury's size and extent. By combining this classification with a recommendation system that suggests hospitals specialized in treating those injuries, the system ensures prompt and proper medical care. The system has been tested against several machine learning models, with GAN achieving 97.92% accuracy in injury classification. This GAN-based approach significantly improves upon traditional methods, potentially saving lives by speeding up the medical response to accidents.