A Study And Realization Of Database Marketing System Of Hongta Group Based On Data Mining

In the era of big data, companies have committed to find an effective way to take advantageof huge amounts of data, in order to make it a powerful tool in competitions between enterprises.Hongta Group has experienced several years of technology constructing and has accumulated alarge number of the business data. Now, the collecting of customer data is also included in thedaily marketing plan. Meanwhile, the Group put forward a strategy that establishing marketingprogram based on analysis result of massive data. But, as public goods, customers of cigarettedistribute quite widly, as the result collecting customer data directly is quite difficult.This paper proposed the building program of Hongta Group database marketing system,achieved the goal of collecting customer data directly, established the interactive platform forbusinesses and consumers, and implemented the data mining of massive customer data andbusiness data, so the Group can take personalized marketing plan for different customers basedon customer analysis and establish scientific market response plan based on business analysis.Since the system completed and put into operation, the customer database had informations ofabout270,000customers, and most of these customers are loyal customers. The research contentof this paper mainly includes the following several aspects:1.Propose the system target based on the marketing strategy of Hongta Group. Get the workprocess of the system according to the system target. Propose the modules that compose thesystem and the functions of the modules according to the work process. The main work processof the system: Collect customer data,integrate business data and store the data uniformlyâ†'Simplify the massive dataâ†'Data mining, establish data model and test the resultâ†'Provide theresult of data analysis for operational staff.2.Design the collection program of customer data. Design the algorithm of a security code,then print the code on the inside of cigarette package, so customers who have bought thecigarette could register and exchange integration with the security code through the clientplatform. The system get the basic information of customers from the register information andthe purchasing information of customers from the security code.3.Design the methods of dimensionality reduction and accumulation of massive data to solvethe problems caused by the computing of large-scale data. Design the Map function and Reduce function of dimensionality reduction and accumulation in the distributed computing modelMapReduce, merge data with reducing the density and the accuracy of data to reduce the amountof data involved in the data analysis in the latter part.4.Study and design the method of data mining of customer data. Build classification tree ofcustomer data to distinguish customers as loyal customer, lost customer and uncertain customer.Design clustering analysis method of customer data based on k-means algorithm to further tapthe potential loyal customers from the uncertain customers. Establish promotion response model,design market value function with the property of loyal customers and lost customers to calculatethe market value of uncertain customers. Design the evaluation method of promotion responsemodel to prove that customers with higher market value have higher responsivity and morelikely to develop into loyal customers.so the Group can input more marketing capital forcustomers with high market value.5.Research and design data mining method of business data. Establish sales forecastingmodel of cigarette based on ARIMA. Count the sales of one kind of cigarette over the past fewyears, get the sequence of the sales based on time, then smooth the sequence. Forecast the salesof this cigarette in the next few months with the forecasting model and the smoothsequence,compare the predicted values with the actual value to prove that the forecasting modelcan predict the sales of cigarette well.6.Propose the technical architecture and deployment scenarios of the system. Implement theload balancing of web server and the expansion of the database server.Cooperate the reverseproxy server with the practical server to implement the load balancing of web server and get theoptimal scheduling policy with several stress tests. Separate the read operation and the writeoperation of the database server to solve the problems caused by the high concurrency of clientplatform.Depart the database on the vertical and the horizontal direction to solve the problemscaused by the storage of massive data.