Short-Term Traffic Flow Prediction Based on the Intelligent Parameter Adjustment K-Nearest Neighbor Algorithm

摘要

Short term traffic flow prediction is important as it supports proactive traffic control and management. A new intelligent parameter adjustment k-nearest neighbor algorithm (IPA-KNN) is proposed for the short-term traffic flow prediction. Two fundamental parameters, the nearest neighbor number k and the predicted sequence length n, are adjusted by an intelligent parameter adjusting method. In addition, the distance measure, including the similarity of fluctuation trend is adopted. Furthermore, a new error measurement method is designed to test the performance of the proposed IPA-KNN model. Compared with the improved KNN method, which is proposed by Habtemichael and Cetin (2016), the data experiments show that the error is reduced by more than 23%.