Electric Machine Sales Forecast Model Based On Machine Learning

Nowadays,China’s Internet information technology develops rapidly.With the help of the rapid development of Internet technology,e-commerce platforms have gradually become an important core for stimulating the overall consumption level of our country,helping traditional enterprises to rebuild their lives,and developing modern service industries.Small and medium-sized shops spring up after springing up.However,due to the lack of factory buildings and equipment at the initial stage of the development of small and medium-sized stores,the banks believe that the risks associated with these small e-commerce companies are too high,so they are unwilling to cooperate with these small e-commerce companies.In order to increase sales and occupy the market,Jingdong Mall has established its own microfinance company.Jingdong Financial has successively launched three core products,namely Jingbao Bay,Beijing Microfinance and real estate financing,which have greatly improved the financing difficulties and high financing costs of small and micro enterprises.Among them,Beijing Microfinance is a kind of credit-oriented financial product.It has the characteristics of high autonomy of loans,no need of collateral,fast capital,low cost,ready-to-use loans and repayments,and online approval.After the merchant has successfully loaned the Jingdong Financial Platform,the funds will immediately go to the relevant accounts in the store.For the offline,it is necessary to measure and track the business status of the shops for loan business.The forecast of store sales is one of the key links in the credit evaluation of “micro-credit” business.Only by accurately estimating the future sales of stores can we reasonably set the loan quota and increase the utilization of funds.This article builds a forecasting model based on past sales records,product information,and product evaluation information to help predict store sales in the future.The data of 1,606 stores in Jingdong Finance and factors affecting sales from August 3,2016 to April 30,2017 are selected,and the forecast model of future sales is given.This paper first makes a summary of the data for each store for 271 days,and then uses the K-means clustering to classify all stores according to the given variables,roughly divided into three categories.The second step is to study the relationship between variables and sales for the stores in each category,and select the appropriate model to fit the data to the prediction model.In this paper,the GBDT algorithm is used in the two types of products,which are the most and the most frequently used,to fit the prediction model.For the category with the least variety of products,the BP neural network model was used to fit the model.The experimental results show that using the above model predictions to obtain better prediction results shows that the model has higher prediction accuracy.It has certain reference value for e-commerce companies of different sizes to formulate marketing strategies.