Data-driven Sales Data Analysis And System Design Of Supermarket Chains

For the traditional retail industry,sales is the key factor that determine its success.The rapid development of the Internet in recent years has led to an explosive growth in the scale and variety of sales data.The decision-making mode that relying on traditional human analysis of massive sales data is increasingly unable to meet the high-efficiency requirements of the enterprise.So it is inevitable to apply data mining technology to sales analysis.Therefore,sales anomaly detecting and locating model and sales forecasting model for the sales of supermarket chains were proposed in this paper and implemented in Spark based on Hadoop.Finally,a prototype system where displaying related data and results was designed and developed.It provides data support for efficient decision making.Sales data are affected by many factors such as holidays and lose comparability in a certain.For supermarket chains,it is an important requirement to detect anomalies,locate anomalies and to realize responsibility to person finally.Thus,it has become a difficult problem to detect and locate anomaly of sales data.Therefore,the sales anomaly detecting and locating model is proposed.Firstly,the model divides the data into four levels according to the management mode of chain supermarkets.Then the original sales data curve of each level is transformed into the weight curve,which makes the data comparable in a certain,and then detect the anomaly and realizes the abnormal location through the establishment of the probability model,and finally realizes the responsibility to person.This model is innovative in anomaly location.Combined with the historical sales data of BBK,the model has some advantages in the accuracy and recall of anomalies.It provides data support for efficient decision-making of enterprises.Regardless of the size of the company,the impact of sales forecast on its logistics,inventory,marketing,finance and so on is extremely significant,but with the advent of big data age,sales forecasting methods based on traditional database technology and experience are clearly difficult to meet the needs of enterprises.Therefore,sales forecasting model based on traditional sales forecasting theory and data mining technology was proposed in this paper.In addition,considering the impact of holidays on sales,the model also deals with the phenomenon of mass consumption in ordinary holidays and the release of spending power before the Spring Festival to improve forecast accuracy.In the BBK 2017 sales forecast,compared with the moving average method,the sales forecast model of this paper has some advantages in terms of MAPE.The visualization of sales data and model processing results make enterprises understand the relevant sales situation at a glance and realize efficient decision-making.