Spatiotemporal Traffic Flow Prediction Using CNN-LSTM Architectures

In this paper, we present a spatiotemporal model for traffic flow prediction using a CNN-LSTM architecture. The proposed method integrates multivariate inputs, including position, velocity, and traffic light states, into a deep learning framework. Our approach leverages the simulation capabilities of SUMO to generate realistic traffic data, which is then used to train the CNN-LSTM model. We evaluate the performance of the model using standard error metrics such as Mean Squared Error (MSE) and Root Mean Squared Error (RMSE). The results demonstrate the effectiveness of the CNN-LSTM model in capturing the complex dynamics of urban traffic and its potential for real-time traffic prediction applications in smart cities.