Research On Dynamic Pricing In Service-oriented Product Revenue Management

With the economic environment changes,the traditional pricing methods have more and more loopholes in the practical application:cannot predict consumer’demand changing accurately,cannot use precision pricing to tap potential consumers,cannot achieve the maximum profit and so on,which have become the several hard problems in enterprise revenue management.We already have studied the service products for the first to find out the characteristics of service-oriented product and pricing model,and then,we summarizes the dynamic pricing literature.After inducing all factors of dynamic pricing and establishing dynamic pricing model we choose "The Reinforcement Learning" what are a popular method in currently and consistent with multi-dimensional characteristics to solve our model problems.Through MATLAB programming we get the optimal pairs of action and state of the model to get the maximize profits.In order to verify the feasibility of model and the final result we classify our case according to whether the service-oriented products moved to do the data simulation and analysis.In the case of once used service-oriented products,we established our dynamic pricing model for hotel housing on the background of hotel housing revenue management.This model combined the actual situation with the remaining number of hotel houses,consumer strategy behavior,pre-sale and previous pricing to do the price decision,then to seeking to the maximize profits.In the case of reusable service-oriented products,we established our dynamic pricing model for high-speed railway dynamic pricing on the background of its revenue management.In the same way,we combined the actual situation with high-speed railway station interval,the remaining number of seats,consumer strategy behavior,lead time and previous pricing to do the price decision,then to ensure high seats utilization and corporate profits.When seeking the optimal solution in two case,we use the popular method of Reinforcement Learning and Q-learning to transform the hotel revenue management and dynamic pricing problem to the state and action process.After agent iteration with Q-learning,we can get the maximum(state,action)to guide the pricing behavior directly of enterprises.After that,we compare dynamic pricing method with traditional pricing method,and analyze the sensitivity of main parameters of the model in the same time,like marketing shares,inventory,cost,price elasticity and so on.Finally,we summarize this paper from three directions:dynamic pricing model,dynamic pricing method,and static sensitivity analysis.Furthermore,we come up with seven conclusions and put forward the corresponding suggestions about revenue management and dynamic pricing in paper.