摘要： 非参数回归方法是短时交通流预测常用的方法,但现有非参数回归方法存在预测速度与精度之间的矛盾。为此,提出一种适用于海量历史数据、基于MapReduce与遗传算法的非参数回归短时交通流预测方法。通过引入MapReduce并行计算框架,加快K最近邻算法的搜索速度。在数据预处理阶段利用遗传算法优化关键参数的设置,并采用MapReduce加速参数优化过程,以解决遗传算法迭代运算时间长的问题。实验结果表明,该方法在保证交通流预测精度的前提下,明显提高了预测速度,并且具有较好的可伸缩性。

关键词: 交通流预测, 非参数回归, K最近邻搜索, 遗传算法, MapReduce编程模型, 并行计算

Abstract: Non-parameter regression method is widely used in short-term traffic flow forecasting,but there is a contradiction on forecasting accuracy and computational efficiency in that method.This paper proposes an improved short-term traffic flow forecasting method based on MapReduce and genetic algorithm in the context of massive historical data.To improve the search speed of K Nearest Neighbor(KNN),a parallel computing framework MapReduce is used to search the KNN.In data preprocessing stage,genetic algorithm is used to optimize the selection of key parameters,and it accelerates parameter optimization process based on MapReduce to solve the problem of long iterative operation time for genetic algorithm.Experimental results show that the method has high scalability,and it can increase the searching efficiency significantly while the forecasting accuracy is guaranteed.

Key words: traffic flow forecasting, non-parametric regression, K Nearest Neighbor(KNN) search, genetic algorithm, MapReduce programming model, parallel computing

中图分类号: 

• TP311