关键词: 非参数回归, 短时交通状态预测, KNN算法, 层次化对象, 自重复性

Abstract:

In order to resolve the inefficient of the nonparametric-regressive model for short-term traffic state forecasting based on KNN(K-Nearest Neighbor)algorithm, the paper presents a data optimization strategy of KNN algorithm. Using time and space characteristics of traffic state, the author constructs traffic state vector with hierarchical object structure, and compresses the historical sample database because of the self-repeatability of traffic state. Experiment shows that the optimization strategy of database improves the efficiency of KNN algorithm.After the compressed data access time is shorter 8.66% than the pre-compressed.

Key words: nonparametric-regressive, short-term traffic state forecast, K-nearest neighbors(KNN) algorithm, hierarchical object, self-repeatability