Abstract
Traditional traffic monitoring systems to detect traffic accidents have many problems including high cost to deploy sensors, limited coverage, and data dependability. The purpose of this study is to use social media data to detect road accidents in real time by monitoring tweets. The proposed method uses Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to evaluate and categorize tweets into three categories: Traffic Accident Tweet, Non-Traffic Tweet, and Traffic Information Tweet. This paper compares the proposed hybrid CNN-LSTM model with other Natural Language Processing (NLP) models such as Logistic Regression (LR), Naive Bayes (NB), Support Vector Machine (SVM), and Random Forest (RF) in terms of accuracy, loss, precision, recall and F1 score. The hybrid CNN-LSTM model demonstrated the highest accuracy in detecting accident tweets, achieving an accuracy of 98.5%. The model also showed low loss, indicating reduced error and improved model performance. These results show how well the hybrid deep learning model uses social media to predict traffic incidents accurately in real time. The study emphasizes how social media analytics and artificial intelligence (AI) may be used into traffic monitoring systems to improve emergency response and road safety. The proposed approach provides a more reliable, more affordable and scalable substitute for conventional sensor-based systems.
Keywords: Accident Detection, CNN, LSTM, Traffic Accident, Traffic Tweet, Tweet Classification.