| From 2016 to now,the revenue growth of Alibaba and Jingdong has dropped to 40%to 50%compared with the previous year,which shows that the trend of e-commerce revenue tends to be flat,and the high-speed growth brought by traffic and mobile Internet dividend has come to an end.So now the e-commerce industry has entered the stage of refined operation,how can better customers need to be more changeable,and the competition among enterprises is fierce.Through the application of new technology to improve the operation level of enterprises,there has been a consensus in enterprises and society,especially in data science and technology.Data mining technology has been applied to the world’s leading enterprises since the beginning of the 21st century.Around 2012,domestic enterprises began to attach great importance to it,but at that time,it was constrained by the lack of a large number of data.After years of data platform construction and accumulation,many enterprises have rich historical data at present.It can be said that the theory,tools and raw material accumulation are basically mature.Enterprises need to effectively forecast many data within their business scope and operation scope,so as to arrange the next stage of production,logistics,marketing,and new business expansion.Through the study of literature and the investigation of enterprises,it is found that the biggest problem in the application of prediction technology in enterprises is the disconnection between the prediction demand of business and prediction technology.Therefore,accurate prediction is the focus of all companies and enterprises.Only by making good prediction can we occupy a strong position in the whole process.The research object of this paper is women’s clothing sales.According to the survey,women users,as the main body of online shopping,account for more than 70%of online shopping.Among them,women’s clothing sales are regarded as the most important part of e-commerce.Therefore,accurate prediction of women’s clothing sales,which accounts for the largest proportion of sales,can reduce operating costs and prepare in advance.The topic of this paper comes from self selected topic.In this paper,three most commonly used data mining models,linear regression,Lasso regression,ridge regression and xgboost,are established by using the exponential smoothing method of time series,ARIMA model and causal prediction method.In addition,the feature of time lag combined with time series is selected for Feature Engineering.Analyze and forecast the data of an e-commerce to establish a complete model.The prediction accuracy of each model is compared to provide effective suggestions for the company in using the model and prediction process. |
