Predicting The Customer Flow Using Machine Learning

The diversification of the economy is booming.Shops face the market which cannot be predicted based on the personal experience.If they can make accurate predictions about the customer flow,the most appropriate decision to respond to the different markets will be made.Therefore,this thesis takes the customer flow of shops as the predicting subject,and carries on the following research.Firstly,according to the characteristics of the customer flow,this thesis analyzed the environmental factors and the product factors affecting it,and studied the influence of different factors.Then the thesis established the theoretical basis for the selection of the predicting data and use the machine learning to learn the data so as to propose the most appropriate data classification method.And it theoretically analyzed different methods of the prediction of the customer flow.Secondly,to meet the demand of the prediction of the customer flow,the program to predict the flow was written,and the method to improve the result was proposed according to the analysis.Still there was a certain modeling error when analyzing the robustness.The analysis and the experiment on the problem of the vanishing gradient and the exploding gradient lead to certain conclusions.Finally,the thesis studied the implementation of the BP neural network to predict the customer flow.The study includes the modeling steps,the optimization of network parameters and the tuning of the parameter adaptation.Then the BP neural network model was implemented in the context of the prediction of the customer flow.The model was compared with the method from statistics.And it proved that predicting the customer flow using the BP neural network was feasible.