The Demand Forecast Of Regional Take Out Order Based On GBDT Algorithm

With the development of Internet plus,people's various lifestyles have changed dramatically.The "lazy economy" has become an economic phenomenon.The combination of Internet and catering industry has made take out a mainstream lifestyle.By forecasting the demand of take out orders that have not occurred in the region,the horseman scheduling can be completed before the order occurs,which makes the intelligent scheduling of the delivery system of the take out platform become a reality,it can effectively improve the delivery efficiency and competitiveness of the take out platform.In view of the above problems,this thesis propose a regional take out order demand prediction model based on GBDT algorithm,which can effectively predict the order demand of every business district in the next hour,and provide the basis for the intelligent dispatching of the delivery system of the take out platform.The main contents of this thesis include the following parts:(1)The thesis introduces the theoretical knowledge involved in the research.It includes the related concepts and steps of demand forecasting and feature selection,and elaborates the algorithm theory of GBDT.(2)The construction of regional take out order demand forecasting model.First of all,the original data is preprocessed.Then,the recursive feature elimination method and cross validation method are used to select the features related to the regional take out order demand prediction,so as to prepare the data for the subsequent model training.In order to improve the performance of the model,it is necessary to adjust the parameters.The manual parameter adjustment work is heavy and subjective,easy to miss,obviously not the best choice.Therefore,Bayesian algorithm is introduced to optimize the parameter adjustment of GBDT algorithm model.Finally,based on the GBDT algorithm after parameter adjustment optimization,a regional take out order demand prediction model is constructed.(3)The thesis uses the order data of a take out platform in Dalian to verify the prediction model and compares the prediction results with the BP neural network and SVR algorithm.The experiments show that the prediction result of GBDT algorithm is better,and GBDT algorithm can effectively predict the demand of the regional take out orders.