本文選題：數(shù)據(jù)挖掘 + 客戶流失預測�。� 參考：《西安電子科技大學》2012年碩士論文

【摘要】：隨著計算機技術(shù)特別是數(shù)據(jù)庫技術(shù)的廣泛應用，各行各業(yè)都積累了海量的數(shù)據(jù)。為了消除“數(shù)據(jù)爆炸但知識貧乏”的現(xiàn)象，數(shù)據(jù)密集型企業(yè)越來越認可數(shù)據(jù)挖掘的重要性。 基于數(shù)據(jù)挖掘的電信客戶流失預測這項研究，是在數(shù)據(jù)倉庫和數(shù)據(jù)挖掘技術(shù)迅速發(fā)展的基礎(chǔ)上，針對電信企業(yè)客戶關(guān)系管理的迫切需要，為消除客戶流失給運營商帶來的不利影響而提出的。 本文的工作基于Q省移動“客戶流失預測系統(tǒng)”項目背景，將數(shù)據(jù)挖掘技術(shù)應用到電信企業(yè)的客戶流失預測中。以Q省移動客戶數(shù)據(jù)、業(yè)務數(shù)據(jù)為依據(jù)，按照商業(yè)理解、數(shù)據(jù)理解、數(shù)據(jù)準備、建立模型、模型評價、模型發(fā)布的步驟，利用統(tǒng)計分析軟件SPSS及數(shù)據(jù)挖掘工具Clementine設(shè)計和建立了電信客戶流失預測的綜合模型。 本文首先介紹了數(shù)據(jù)挖掘理論及分類預測算法，并詳細描述了生存分析理論及比例風險模型。在建模過程中，重視數(shù)據(jù)質(zhì)量，進行了有效的數(shù)據(jù)清洗、轉(zhuǎn)換、探索工作，，處理了不平衡數(shù)據(jù)集，并在業(yè)務經(jīng)驗及屬性約簡的基礎(chǔ)上建立了流失預測指標體系。最終建立了決策樹、神經(jīng)網(wǎng)絡(luò)、logistic回歸以及生存分析Cox模型，并對模型進行了多項指標的評估。在維系挽留工作中，初步分析了客戶流失原因并評定了客戶價值，提出針對不同客戶進行因時因地的有針對性的維系挽留策略，減少挽留成本并提高挽留的成功率。 本文把數(shù)據(jù)挖掘理論與實際項目相結(jié)合，建立了基于數(shù)據(jù)挖掘技術(shù)的電信客戶流失預測綜合模型。理論研究上對分類模型及Cox模型的構(gòu)建具有指導意義；應用的結(jié)果表明所建立的模型能夠給決策人員提供有價值的預測信息并給出相應的解決方案。
[Abstract]:With the wide application of computer technology, especially database technology, huge amounts of data have been accumulated in various industries. In order to eliminate the phenomenon of "data explosion but poor knowledge", data-intensive enterprises increasingly recognize the importance of data mining. The research of telecom customer churn prediction based on data mining is based on the rapid development of data warehouse and data mining technology, aiming at the urgent need of customer relationship management in telecom enterprises. In order to eliminate the negative impact of customer drain on operators. Based on the project background of Q province mobile customer churn prediction system, this paper applies data mining technology to customer churn prediction of telecom enterprises. Based on Q province mobile customer data, business data, according to business understanding, data understanding, data preparation, modeling, model evaluation, model release steps, Using the statistical analysis software SPSS and the data mining tool Clementine, a comprehensive model of telecom customer churn prediction is designed and established. This paper first introduces the theory of data mining and classification and prediction algorithm, and describes the survival analysis theory and proportional risk model in detail. In the process of modeling, we attach importance to data quality, carry out effective data cleaning, transformation, exploration, deal with unbalanced data sets, and establish a loss prediction index system on the basis of business experience and attribute reduction. Finally, the decision tree, neural network logistic regression and survival analysis Cox model are established, and several indexes are evaluated. In order to reduce the cost of retention and improve the success rate of retention, this paper analyzes the reasons of customer turnover and evaluates the value of customers, and puts forward a targeted retention strategy aimed at different customers in order to reduce the cost of retention and improve the success rate of retention. This paper combines the theory of data mining with practical projects, and establishes a comprehensive model of telecom customer churn prediction based on data mining technology. The theoretical research is of guiding significance to the construction of classification model and Cox model, and the application results show that the established model can provide valuable prediction information and corresponding solutions for decision makers.
【學位授予單位】：西安電子科技大學
【學位級別】：碩士
【學位授予年份】：2012
【分類號】：TP311.13;F274;F626

【相似文獻】

相關(guān)期刊論文 前10條

1 張韜;胡e

本文編號：1789480