FastGCN Framework for IoT-Powered Accident Detection and Emergency Response

More cars on the road means more daily accidents. Medical and rescue delays are maj or causes of death in such incidents. A cognitive agent-based smart accident alert and rescue system could save lives by reducing delays. This concept fits inside intelligent transport systems (ITS), which are known to improve traffic safety, especially in smart cities. The proposed system includes preprocessing, model training, and feature selection. To ensure data quality, preprocessing cleans the dataset of missing values, noisy data, outliers, and normalisation. Structure-From-Function-Movement (SOFM) was used to find and remove duplicates, optimising the dataset. The FG-Convolutional Recurrent Neural Network (FG-CRNN) was used to train the model because it can capture complex temporal patterns in accident data. Our method beat CNN and RNN models. This model's average accuracy of 95.86% shows its reliability and efficacy in real-time accident prediction and alert systems. This cognitive agent-based system could enhance accident rescue reaction times, making ITS in smart cities viable. The FG-CRNN model's accuracy shows its usefulness and stresses the importance of sophisticated feature selection and data preparation for optimal results.