Research And Application Of Decision Tree In Mobile Marketing

The battle about users in mobile telecommunications markets has beenintensified. Especially in the arrival of4G era, mobile telecom operators doeverything they can to obtain more users. Group customers of mobiletelecommunications users as a strong social impact can bring huge profits to largecustomers for telecom operators, which is the focus of the battle. Meanwhile, theoperating system of mobile telecom operators retains a great deal of users’information data. These data containing a wealth of customer information caneffectively analyze users’ needs and satisfaction for the service through data miningtools. Operators can use this information to develop an effective marketing strategy toachieve victory in the battle. Decision tree algorithm in data mining technology iscurrently a very popular classification prediction algorithm. Decision tree algorithm isan inductive learning algorithm based on examples, reasoning a classification rulerepresented in a tree structure through learning the training sets. Decision treealgorithm becomes one of the most widely used techniques in data miningapplications because of its outstanding data classification efficiency and intuitivedisplay of the results.Based on the study of decision tree data mining technology, the thesis analyzesand digs mobile user data of group customers. Potential false mining system ofmobile group customers is a visual mining platform proposed by mobile company inorder to implement the group customers’ meticulous management and enhanceoverall level of customer service and marketing, whose core is the classifier based onID3algorithm of decision tree. The members of the group customers are classifiedinto potential members, false members and sticky members by the classifier. However,there are two drawbacks in ID3algorithm. One is that ID3algorithm calls a largenumber of system function to calculate logarithmic function. The other is that whenselecting the best classification property, ID3algorithm prefer the attributes whosevalue is the most, but they often have a relatively low correlation with businessproblem. First, the thesis simplifies ID3algorithm by McLaughlin formula,transforming a large number of log function into simple multiplication and divisioncalculations and improving the efficiency of the algorithm. At the same time, according to the own characteristics of business attributes of group customers,attribute viscous degree is proposed. The attributes in the candidates attribute sets aresorted by calculating the viscosity of attributes. When selecting the attributes, theproperty whose viscosity value is higher is preferred, effectively avoiding biasproblem exists in ID3algorithm. Finally, the improved ID3algorithm is applied topotentially false mining system of mobile group customers. The classification resultsare shown by Java Web technologies of SSH framework and foreground statementsgraphics display technology based on JQuery and Fusioncharts.