Research On Sales Forecasting Method Based On Time Series Analysis

With the development of Internet and information technology,and the rise of cloud computing and big data technology,it is possible to implement the artificial intelligence.Now the standard of people's living is better than before,and sales is an eternal topic.Mobile payment and online transaction have brought great convenience to our life.The information management ways of data management have greatly improved the work efficiency of the enterprise,and during the time accumulated a huge amount of data.Nowadays,people have been able to recognize the importance of data and the information included in it.Sales data is a class of time series data.If we can use the great model algorithm to describe its pattern,we can analyze and predict its future trend and give the estimated value.For the enterprise,this is very important.According to the forecast results of product sales,companies can develop more accurate marketing strategies to increase sales profits,reduce costs,and even avoid losses due to inaccuracy in the market.In this paper,we aimed at the phenomenon which is called "data explosion and lack of knowledge" in real enterprises,the time series mining of sales data is studied.Sales data can be obtained through crawler technology and data provided by enterprise users.The traditional time series analysis methods,ARIMA and Prophet are used to analyze and predict sales data.Based on the ARIMA algorithm,the improved model algorithm ARIMA+SVR and ARIMA+Boosting are proposed.The sales data are analyzed and predicted,and the experiment is carried out on the real sales data of the enterprise,and the comparison of the experimental results is given.The experimental results show that the improved algorithm is better than the traditional single model algorithm in the efficiency and accuracy of the algorithm,and it is more accurate and reasonable in the trend prediction of time series.Finally,we develop a time series analysis platform based on sales data,building a data platform,and integrating our model algorithm to the data platform to provide users with massive data processing and visualization of time series analysis function modules.