Composite of support vector regression and evolutionary algorithms in car-rental revenue forecasting

In the car-rental industry, it is generally assumed that a 1% increase in price contributes an 8% profit increment. Therefore, pricing is of priority concern in revenue management. Accurately forecasting monthly revenue per unit (RPU) (per car) has received increasing attention owing to its ability to provide a benchmarldng performance index for annual pricing. However, RPU in the car-rental industry suffers serious nonlinearity. Recently, support vector regression (SVR) has been successfully applied to solve nonlinear regression and time series problems. This study elucidates the feasibility of using SVR to forecast RPU. Moreover, the parameters of a SVR model are derived using the immune algorithm (IA). Subsequently, RPU data from DTAG (Dollar Thrifty Automotive Group, USA) are used to illustrate the proposed SVRIA (support vector regression with immune algorithm) model. The empirical results reveal that the proposed model outperforms the other two models, namely the Holt-Winters (HW) model and the seasonal HW (SHW) model. Consequently, the SVRIA model provides a promising alternative method of forecasting RPU.