Forecasting of National Expressway Network Scale Based on Support Vector Regression with Adaptive Genetic Algorithm

Abstract

Forecasting of national expressway network scale is an important part of expressway network planning. It is a complicated problem due to its strong nonlinearity and small quantity of training data. Firstly, An AGA-SVR forecasting model is proposed by combining the support vector regression (SVR) and adaptive genetic algorithm (AGA) in this paper. SVR has been successfully employed to solve regression problems of nonlinearity and problem samples, but selecting appropriate parameters, which is a combinatorial optimization problem, is very crucial to learning performance and generalization performance of SVR. An AGA is used to determine the free parameters of support vector regression. Finally, examples of the scale data of China national expressway network are used to illustrate the performance of the proposed AGA-SVR model. Experimental results demonstrate that the AGA-SVR outperforms the BP neural-network; the AGA-SVR model is a reasonable alternative to forecast national expressway network scale.