Description of Invention
This project investigates the application of computational intelligence in traffic flow prediction. Leveraging neural network architectures such as ANN, DNN, RNN, and LSTM, the study aims to overcome the limitations of traditional models like ARIMA in handling non-linear and dynamic traffic patterns. Using a dataset from 35 traffic routes, extensive preprocessing, hyperparameter optimization, and model training were conducted. The evaluation metrics—MSE, MAE, and R²—highlighted the MLP model's superior performance. Although the models captured traffic trends effectively, challenges remain in addressing extreme fluctuations. Future work will focus on incorporating external factors, like weather data, to improve prediction accuracy.