Enhancing Traffic Forecasting Accuracy with Fuzzy Data Fusion Learning

This article explores a challenge related to managing and forecasting short-term traffic flows. The accuracy of this prediction is the subject of much research because it directly affects the performance of many traffic applications. In particular, ensuring the monotonicity of traffic variables is predominant for accuracy. In this paper, a predictive model-based Fuzzy Traffic Data Fusion (FTD F) system is proposed to predict traffic based on multiple sources. The model incorporates fitted features via fuzzy integral and Multiple Linear Regression (MLR) to ensure monotonicity by assigning importance weights to feature sets. The evaluation step is based on analyzed traffic data, the results of which show a high approximation rate of the predicted data compared to the real data and the prediction error is significantly reduced. Additionally, the performance outperforms existing predictive models using the prediction criteria.