Cloud differential evolution algorithm for traffic flow prediction with least squares support vector machine

Abstract

Differential evolution (DE) algorithm mainly uses the distance and direction information from the current population to guide search. However, it has no mechanism to extract and use global information about the search space. Cloud model is an effective tool in uncertain transforming between qualitative concepts and their quantitative expressions. It can be used to extract the global information about the search space. In this paper, a novel cloud differential evolution algorithm (CDEA) based on cloud model and differential evolution is firstly introduced. In the algorithm's offspring generation scheme, new individuals are generated in the cloud model way or in the DE way. In the cloud model way, three digital characteristics of the best individuals from the current population are firstly calculated by backward cloud generator. And then they are used to produce the new individuals by positive cloud generator. Both the global information from cloud model and local information from DE are used to guide the further search. Then, least squares support vector machine (LSSVM) is used for traffic flow forecast model, and CDEA is used to optimize two parameters of LSSVM. The experimental results show that LSSVM with CDEA gives better accuracy in terms of MSE, MARE and EC for real traffic flow data.