Research Of Credit Of Fixed Telephone Customer System Based On Data Mining Technique

Start in 70's of the 20 century, the database technique turned into the stage of the quick development, each core database of big business enterprise was increasing quickly by the speed of the TB class. The data conceals many important information, people hope as to it's can carry on the analysis of the higher layer, for the purpose of better utilization of these data . The traditional relation database system can carry out a data record ,search ,covariance etc., but can't discover the existent relation and the rule in the data, can't prepare future trend according to the existing data. In a word, the IT industry is widespread to lack the means with knowledge that the excavation data currently, causing the phenomenon of "the data enrich, but the knowledge is needy" . Thus the collections become "the data grave" in the data of the large database system- Rare data file visited again, important decision usually is not according to database information enrich of data, but according to the intuition of the decision maker, because the decision maker lacks to withdraw tool of have the worth knowledge from the amount of sea data. Consider a current expert system technique in addition, usually, this kind of system depends on an expert to input the knowledge into the knowledge base artificially in the customer or realm, this process often has the deviation and error, and more expensive. People expect a kind of efficiently system that can discover the amount of sea data to help the business enterprise and discover bureau to make a right judgment, turn "the data grave" to "the data gold". A kind of database structure that have appear since the end of last century is called the data warehouse, this is a kind of organized store of many pieces according to the source the station, to support management. The data warehouse technique includes a data clean ,the data integration and online analysis processing[OLAP). Though the OLAP tool support multi-dimension analysis and decision, get success in the discover knowledge of the amount of data, but for the analytical function of the deep level, such as the data classification ,gathering and discover the characteristic of data by time, still need other analysis tools. The knowledge in the database discover KDD (the Knowledge Discovery in Database) is to withdraw a deluxe processing process from a great deal of data an authentic and novel of , valid and can be understood by people. Generally the stage of the knowledge study can be called the data excavation in the KDD (Data Mining). The data mining is a kind of technique to withdraws the estimate information that conceal from the large database or the data warehouse. It can dig out the data latent mode, finding out the most worthy information, guiding the business behavior or lending support to a science research. In various academic works and the periodical, for being absorbed in the research of the core to combine convenience to express, in fact already take the "data mining" the righteousness phrase of the knowledge detection in the database. At our country, many discover bureau have already carried on data mining technical research for a long time, but there is no typical model case of applied large system of data mining. However, in the banking and the telecommunication industry ,there is the applied need and premise of the existence of data mining because of the higher information degree. Mainly in the applied realm of the telecommunication ,there is customer relationship management, customer credit degree analysis, the customer run off analysis, the customer consumes mode analysis, market expansion analysis etc.. Because China join WTO, the market environment of the internationalization requests public telecommunication inside the country keep up with the abroad telecommunication in conduct and management to meet the internationalization competition of telecommunication business. Along with the telecommunication market, the center of gravity of the competition is also changed to service, The marketing subdivides request subdivide to the customer, the composing of customer usually decide the success or failure to the establishment of the marketing policy.The telecommunication corporation is a data intensive industry, have already collect and purified a great deal of data through the business of many years data backlog. How to make use of these amount of data, carry on valuation to customer's credit degree, solve the "mind decide" phenomenon in the marketing policy establishment, is the urgent demand of the telecommunication business enterprise.The purpose of this topic is to research the realization of the data mining, and make use of this technique to the customer credit degree of the fixed network telecommunication corporation. Make the telecommunication business enterprise exaltation competition ability, acquire more profit. This text introduced data mining foundation theories firstly, and related calculations. Take actual item as to rely, with the CRISP_DM ( Cross_industry Standard Process for Data Mining)set up mode for frame, gradually according to : business comprehension, data comprehension, data preparation , model establishment, model choice, model release, make use of the data mining software Clementine , carry out the fixed network telecommunication customer credit degree design and the realization.The data mining technique is doubtless established on the foundation of large amount data, in front of large amount of data, a lot of works needed in the aspect of data withdraw , organize , filter etc. this can be certificated in similarly of the item of the data warehouse. Because the characteristics of data source of the our country telecommunication enterprise is complicated, more data noise, the work of this aspect is more important. Therefore on the foundation of the CRISP_DM model, combine physically, the thesis also introduced a concrete data to withdraw, organization process and filter , Enclosed with main procedure and table structure.Through the first step of data organization, we use data mining software Clementine tool to carry on the process of the data quest, then introduces how to set up the model by Clementine. We used the artificial nerve network and the decision tree calculations, building up model, carrying on an estimate, and used the Analysis node to carry on valuation to the model, then carried on a ratio with the actual data, and according to actual result ,we selected the most in keeping with calculations to apply to reality as a result .On the end we use BO statement tool to carry on a demonstration.In the whole data mining process, the data choice ,data clean , establishment ,data integration , foundation of model, model excellent turn is a continuously circulating process. Pass to adjust again and again to this process , observation, then can attain ideal result. Repeat many times to the above-mentioned process is also the characteristics of data mining. Even many possible new circumstances appears after the establishment of the model success, for example the new business release, change of time, need to re-set up a model, so the data mining process need to be repeat continuously.Today, these mature technique, high performance engine of the relation database, let the data mining technique turn into a practical stage in the current data warehouse environment, The data mining technique also put forward many application in the telecommunication currently. How to combine the current present condition of the telecommunication enterprise, make use of this new technique availably in business enterprise, build up accurate data model, and carry on the usage model estimate, produce results quickly, is the meaning of this topic research.