Customer Identification, Transaction Generalized Calculation And Classification And Regression Tree-based Telecommunications Industry

Along with the rapid development of the technology of data warehouse and data mining,customer relationship management (CRM) becomes more and more important. On this need,we advanced the project of abnormal customers recognition system based on generalized computing and classification and regression trees (CART).Based on the generalized computing theory,The thesis combines multi-rules neuralnetwork with a kind of decision tree-classification and regression trees. Furthermore,we put forward a new kind of abnormal customers recognition model. The model can improve classification precision and recognition efficiency effectively,make full use of the advantages of multi-rules neural network and classification and regression trees,and make up their respective disadvantages at a certain extent.The work that is carried out by me for this project as follows:At first,works over the decision tree technology and the multi-rules neural network theory based on the generalized computing,outlines the advantages and disadvantages of the two theories,analyzes the possibility to combine multi-rules neural network with classification and regression trees,and studies some achievement in this field.And then,the thesis brings forward a new modeling method-abnormalcustomers recognition system based on generalized computing and classification and regression trees. The system is composed of multi-rules neural network learning part and classification and regression trees processing part. Multi-rules neural network learning part decreases the dimensions of attribute collection,to reach the goal of simplifying the input;we stress the multi-rules learning algorithm based on fuzzy entropy rule;at the same time,all the knowledge available is used to design the input layer,hidden layer and output layer of the neural network. Classification and regression trees processing part introduces growing algorithm of CART,pruning algorithm of CART and selecting best tree algorithm etc.On the basis of the concerned new model,the thesis presents in details the designing of multi-rules neural network based CART system for abnormal customersrecognition. An concrete application embodies the superiority of the new method advanced in the thesis.At last,the thesis summarizes the work that has been done,indicates the problems that have not been resolved and describes the prospect of the future task.The thesis combines generalized computing theory with classification and regression trees technology,makes the great theory innovation. The final project is applied into abnormal customers recognition,the result shows that it has immense application value. The innovation of the thesis as follows: Advances the generalized computing theory that combines symbolic computingwith neural computing,fuzzy computing and evolutionary computing. Brings forward the multi-rules neural network learning method based ongeneralized computing theory. Improves on the classification and regression trees technology,increases it'sclassification precision. Combines multi-rules neural network with classification and regression treestechnology based on generalized computing theory,implements the abnormalcustomers recognition system.Compare with the traditional CART recognition system,the concerned model in the thesis has the following characteristics:Neural network learning method based on fuzzy entropy rule classifies more accurately,converges more rapidly and computes more quickly;Getting rid of redundant information,simplifying the attribute collection;The result of the concerned model can be understood easily;Making lull use of the two theories,avoiding their respective disadvantages.