| In recent years, the data excavating has been a hot spot in research field of the database. The data excavate technology is used in sales forecast and analysis of the retail business, analysis of the credit of customer of financial circles and analysis which the customer swindle, value analysis of the customer of the telecommunication industry sales forecast, ect. But because of the development and management of the above, trade are still at extensive stage. There is certain disparity with foreign countries. Customer's basic materials are being perfected progressively. Analysis and prediction of utilizing the data excavate technology are still care.In the field that telecommunication characteristic analysis and analysis of customer's value of utilizing the data excavate technology are still black what is the thesis is studies is the model of subdiving how to get the customer of the telecommunication industry through the cluster. The model can be known customer with similar characteristic group , thus help to understand the customer. It will become and analysis the customer and take shape on the market tacties finally. In appropriate time, through the appropriate channel, it can offer appropriate services to appropriate customer really, in order to meet customer's needs.The text has carried on more overall comparative research to the basic method-cluster's technology. There are main research work and characteristic:l)The text has put forward a kind of new level cluster's method. It receives the inspiration of CURE algorithm. It adopts a lot of central points to express everyone effect,. Different from algorithm of CURE, the algorithm of this text adopts the method to divide to divide the data into the atom cluster at first. The based on these atom clusters, implement bottom-up level cluster's receiving the field results. The method of this text can not merely be discerned kind of any form size, can fitter outside the noise data effectively, time bottom very too complexity.2)Do well on the low data cluster's quality drops which dealing with the high data of linking. From meet telecommunication data amount of trade heavy, link bottom high characteristic, receiving CLUQUE inspiration research work, It has adopted the spare model, has put forward a kind of practical and high-efficient cluster model. Algorithm, bases on the analysis with space in it, it need a sure parameter in advance to avoid making great efforts, performance of efficiency and information processing that improve in spacetime at the same time. Through a chapter has got the inspiration of CLUQUE algorithm, this chapter has proposed except algorithm has a lot of differences with CLUQUE algorithm, carry on it beyond greater improvement have a lot of some innovative contents, impesent research work both at home and abroad oneself. Algorithm that this text put forward is adopted top-down with the bottom-up way that combines together. Collect the whole sample as every one first, in some coordinates are linked, the level cluster using this text to put forward carries on preliminary division on the whole sample. Then according to it is carry on further division to these kind to confirm to link while being other, produce it to the asymmetrical level cluster of part of introduction sample finally. Secondly, it needn't appoint parameter value in advance in algorithm that this text put forward .Solve and threat cluster of different density disern less questions of kind of the quantity of samples, the complexity of algorithm is lower than CLIQUE algorithm.3) It succeeds in applying the model of this point of clusters to the customer of telecommunication, industry the model of subdiving. Through excavating the data to relevant attribute, such as user's materials, conversation behavior, service act etc. Analysis the conversation behavior characteristic and service type and each age bracket user's night behavior characteristic of each age bracket users. The experiment result has verified the validity of this cluster model.
|
PDF Full Text Request