WKNN-FDCNN method for big data driven traffic flow prediction in ITS

Traffic prediction is a vital paradigm in intelligent transport system (ITS) due to the increase in traffic flow. The big data traffic flow prediction faces heterogeneity and complexity in data samples due to the huge number of data samples. The proposed WKNN-FDCNN method simplifies the big data handling process by utilizing the Weighted K Nearest Neighbour (WKNN) algorithm for data mining and a Fuzzy based Deep Convolutional Neural Network (FDCNN) for prediction. The spatio-temporal characteristics of traffic flow data are modeled as a weight function in WKNN, which helps in handling heterogeneity and complexity in data samples. The fuzzy logic incorporates uncertain information from real traffic flow data to improve the prediction performance. Finally, a DCNN approach is designed to predict the traffic flow using spatio-temporal features, traffic state information mined using the WKNN, and fuzzy traffic rules. The WKNN-FDCNN outperforms the conventional approaches in terms of Root Mean Squared Error (RMSE= 13.27), Mean Absolute Error (MAE= 10.34), R-square (0.98), and Mean Absolute Percentage Error (MAPE= 0.92) in the PeMSD4 dataset. The proposed method contributes to the development of intelligent transportation systems and provides a promising solution to handle big data challenges in traffic flow prediction.