Travel Time Estimation Method Using SCATS Traffic Data Based on k-NN Algorithm

A new method was designed to improve the effect of travel time estimation using SCATS traffic data. In this method, by analyzing the characteristics of SCATS traffic data that their acquisition intervals are not strictly consistent, the virtual time series of SCATS traffic data was constructed first. Then, factors of cumulative squared loading over 85% extracted by factor analysis were included in the traffic state feature vector, and the Euclidean distance was used to measure the closeness between current traffic state and historical traffic state. Finally, the number of nearest neighbors that correspond to the minimum error of travel time estimation was selected, and the weights of k-nearest neighbors were identified by normalizing the reciprocal of the distance between traffic states, based on which the travel time estimation method was established. The results indicate that compared with the multiple linear regression (MLR) method, the proposed method can reduce the mean absolute error (MAE), mean absolute percentage error (MAPE) and root-mean-square error (RMSE) of travel time estimation by an average of 9.68 s, 8.07% and 4.5 s, respectively.