Unconstrained Car Rental Demand Forecasting Model Based On Revenue Management

Chinese car rental industry has been developed for nearly 30 years,and the simple business model has been unable to adapt to the rapid development.The companies start to explore the application of revenue management.Capacity control,oversold and pricing strategies can increase the expected return,but it depends on the accuracy of demand forecast.The demand forecast is based on the historical demand data,but the historical data is often constrained,that can not reflect the real passenger demand situation.So it is need for constraint data without constraint repair.Using of repaired unconstrained data for future time demand forecasting is called unconstrained demand forecasting.Due to the high number of vehicles in the car rental site,the high subjective initiative,and the high probability of upgrading and downgrading,the unconstrained demand forecast becomes a trouble of revenue management.The unconstrained demand forecast of car rental mainly includes unconstrained recovery of historical constraint requirement and demand forecast of future time.The main work and research results are as follows:First of all,the paper analyzed the development situation and the conditions of revenue management with the study of car rental revenue management as the research object based on the literature review.It was important to obtain unconstrained demand forecast,based on the characteristics of car rental revenue management.Secondly,an unconstrained demand repair method was put forward that was taken into account customer choice behavior.The traditional unconstrained demand restoration method was based on the traditional revenue management,that could not describe the subjective choice behavior of the customer.The customer car rental intention survey was designed,and used the analytic hierarchy process of variable precision rough set to deal with the data,and got the factors and weight.Customer satisfaction survey was designed,and used MNL to process the data Customers choose the probability of preference to improve the traditional Spill model.Thirdly,a combined forecasting model was put forward based on linear and nonlinear method.Holt-winter model had a better forecasting effect on the seasonal and cyclical changes in car rental demand,but the customer demand was a complex sequence affected by many factors.The linear prediction could not meet the requirements of increasing accuracy.The nonlinear BP neural network had strong fault-tolerant ability and associative memory ability,but it had the risk of falling into the local optimum.Therefore,the two methods were combined reasonably,and the useful information of each individual model could be used to avoid the error caused by the single prediction.Finally,through the numerical examples,the traditional repair model and the improved model of the unconstrained demand repair results were got.By using the unconstrained demand after repaired,three forecasting methods were needed to predict the demand for the next cycle.Compared with the real demand,the feasibility of the combined forecasting model and the improved Spill model considering customer selection behavior were verified.