Research And Application Of Decision Tree Algorithm In CRM Of Tobacco Company

With the increasing of national tobacco control,tobacco company is facing unprecedented operating pressure.Cigarette retail customers is the bond between consumers and tobacco system.The potential customers in these retail customers plays a very important role for tobacco company business and profit.How to select potential customer from retail customers is an urgent problem to be solved in the customer relationship management of tobacco company.Inner Mongolia kunming cigarette limited liability company accumulated a lot of customers data.Based on the analysis of these data,Inner Mongolia kunming cigarette limited liability company built the potential customer classification decision tree model using decision tree algorithm and developed the customer relationship management(CRM)system using the classification rules extracting from the above-mentioned decision tree model.The main research work of this dissertation includes:(1)Analyzed the operation and development present situation of the tobacco system.Analyzed the Data Mining-Customer Relationship Management(or DM-CRM for short)system requirements,design background and way of thinking of the Inner Mongolia kunming cigarette limited liability company.The principle and application of decision tree algorithm in data mining are studied.(2)According to the goal and the demand of data mining in the DM-CRM system of Inner Mongolia kunming cigarette limited liability company,set up the integrated environment for data mining and system developing based on Phpstorm,XAMPP,SQL Server,and DMX language.Selected and preprocessed the customer history data of BaoTou city in 2016 as the model training samples.Based on ID3 algorithm,the decision tree model of the potential customer classification was constructed and the classification effect of this model was verified by the fit evaluation method.(3)Participated in the requirements analysis and part design task of the DM-CRM system of Inner Mongolia kunming cigarette limited liability company.The corresponding classification rules are generated by the potential customer decision tree model that constructed in this dissertation,and the development of the "potential customer management" function module in DM-CRM system is completed by using these rules.