Short-Term Demand Prediction Method for Online Car-Hailing Services Based on a Least Squares Support Vector Machine

Abstract

The purpose of this paper is to study the short-term demand prediction for online car-hailing services problem. Prediction of short-term network car demand can provide many benefits such as an increase in the income of network car drivers. In addition, demand prediction is an important resource for recommendation systems, carpooling systems, and network car scheduling; and therefore, predicting the demand for network cars has great significance. In the last few decades, scholars have studied various problems related to short-term demand prediction for online car-hailing services based on clustering algorithms and regression algorithms. However, these studies are still problematic because the accuracy of demand prediction is not high enough. Therefore, this paper studies a method of improving the accuracy and the efficiency of demand prediction. Due to the high prediction accuracy and the fast training efficiency of least squares support vector machine (LS-SVM), a short-term demand prediction method for online car-hailing services based on LS-SVM is proposed. The modeling process involves selecting the dependent and independent variables, the basic principles of the LS-SVM, the kernel function and superparameter, model training, and prediction. In a numerical experiment, we use network car order data as the network car demand data to test the model. We show that the experimental results with the LS-SVM method implemented in this paper and then compare the model with lasso linear regression, nearest neighbor regression, decision tree regression, and neural network. The experimental results show that the short-term demand prediction model for online car-hailing services based on LS-SVM performs better than the other methods.