Dissertation > Excellent graduate degree dissertation topics show

Study and Application of Large-Scale One-to-One Marketing Optimization Algorithm Based on Predictor-Corrector Method

Author: HeZuo
Tutor: WuMin
School: Central South University
Course: Control Theory and Control Engineering
Keywords: Customer Relationship Management One-to-one Marketing Optimization Predict-Correct Method Bayesian Be-lief Network Customer Behavior Prediction Method Elimination Tree LDL Factorization
CLC: TP301.6
Type: PhD thesis
Year: 2006
Downloads: 245
Quote: 0
Read: Download Dissertation

Abstract


One-to-one marketing optimization is a crucial technology for analyt-ical customer relationship management (CRM). Its main function is assist-ing an enterprise in one-to-one marketing activities to improve customer re-tention, customer loyalty, and profitability by providing suitable decisionswith understanding and influencing customer behavior through meaningfulcommunication. However, the methods based on inference business rules,which are used in existing CRM software, cannot guarantee that enter-prises can gain the maximum profit. If this project is studied by operationalsubject, for a large-scale enterprise, the one-to-one marketing optimizationmodel is very complex and the scale of it is very huge, the conventional lin-ear programming methods cannot solve it. So, large-scale one-to-one mar-keting optimization model, large-scale one-to-one marketing optimizationalgorithm and implementation and application method for this algorithmare studied in this paper, and the main creative contents included in thispaper are the following five aspects.(1) One-to-one Marketing Optimization Modeling MethodSince modeling one-to-one marketing optimization has many difficultpoints, such as diversity of objective function, complexity of constraints,parameter uncertainty and existing some nonlinear conditions etc, this pa-per analyzes one-to-one marketing mechanism, business rules and market-ing experience to construct mulit-objective and muilt-constraint one-to-onemarketing optimization linear model by consideing marketing cost, cus-tomers’ fancy and characteristic of products and channels. This model isuniversalizable, and can indicate one-to-one marketing optimization prob- lem exactly.(2) Customer Behavior Prediction MethodIn order to solve the problem that customer responses probability can-not be predicted accurately in the process of modeling one-to-one market-ing optimization, a new customer behavior based on Bayesian networkprediction method is presented. Though analyzing probability depend-ing among attributes of tables in historic business marketing informationdatabase, it constructs a Bayesian belief network, and does probability in-ference on some prediction instances based on this Bayesian belief networkto predict customer responses probability by calculating the joint probabil-ity. Considering that the scale of database information is very huge forlarge-scale one-to-one marketing optimization problem, a new Bayesiannetwork learning algorithm is presented in this method, whose time com-plexity is O(n~4), where n is the number of nodes, instead of exponentialtype. Compared with other existing customer behavior prediction method,such as based on naive Bayesian classification, this method considers theconditional independence among customers, products and channels, andcan calculate the probability that customer behavior will happen, so it hasmore accuracy.(3) Large-Scale One-to-one Marketing Optimization AlgorithmConsidering that the numbers of constraints and variables are veryhuge for a large-scale one-to-one marketing optimization model and an or-dinary personal computer has limited performance on memory and CPU, alarge-scale one-to-one marketing optimization algorithm with O(nlogn/ε),where n is the number of decision-making variables andεis the tolera-ble error, time complexity and superlinear convergence rate is presented.Based on predict-correct method and combined with some large-scale sparsematrix calculation methods, such as column approximate minimum de-gree ordering, LDL factorization and block gauss elimination etc, it cansolve the large-scale one-to-one marketing optimization problem, which the number of customer is 100 thousands or above.(4) Large-Scale LDL Factorization MethodA new LDL Factorization approach is presented to solve huge lin-ear system involved in large-scale one-to-one marketing optimization al-gorithm based on predict-correct method. It breaks the computations in-volved in LDL factorization down into two stages: the pattern of nonzeroentries of the factor is predicted, and the numerical values of the nonzeroentries of the factor are computed, and does recursive calculate accordingto topical sort of elimination tree of its factorization, so that it can reducememory usage and avoid unnecessary numerical operations, and can solvelarge-scale linear system efficiently.(5) Implementation and Application of Large-Scale One-to-one Mar-keting Optimization AlgorithmCombined with Dynamic-Link Library, multi-threading and memorypool management technology, the large-scale one-to-one marketing opti-mization algorithm is programmed with memory constrained system byobject oriented approach. With the analysis of CRM framework, the appli-cation approach of one-to-one marketing optimization for modern enter-prise is studied and a one-to-one marketing optimization decision-makingsoftware, WHCRM, is developed. This software provides automatic mod-eling function for one-to-one marketing optimization, and the maximal ca-pability is several hundred thousand customers.With the study of one-to-one marketing optimization problem for large-medium enterprises, the centralized management mode for customer, prod-uct and channel information is presented, which can improve employees’work efficiency and positivity; and the concept of optimization can avoidrepeated and useless work by the greatest extent, so it can save marketingcost and resource largely and will be helpful to build economized society.

Related Dissertations

  1. Design and Implementation of Cross-selling System Based on Knowledge Reasoning,TP18
  2. Research on Customer Collaboration-based R&D Management of Inspur Group’s ERP,F426.672
  3. The Design and Implementation of Banking Personal Customer Relationship Management System,TP311.52
  4. The Analysis and Application of Telecom Customer Retention System Based on Data Mining Technology,TP311.13
  5. Design and Implementation of customer relationship management system based on Web,TP311.52
  6. Design and realization of the bank customer relationship management information system,TP311.52
  7. The Research on CRM Applying in Direct Marketing Business,F274
  8. Study and Application of Safety Monitoring System for Smart Home Based on Wireless Sensor Networks,TP273.5
  9. CRM Implementation in Pharmaceutical Industry,F426.72
  10. Call Center Design and Implementation,TN99
  11. Research on Customer Relationship Management in Third-Party Enterprise,F274
  12. Design and implementation of CRM system based on the SaaS model construction machinery sales agency,TP311.52
  13. Research on Relationship Marketing in the Brokerage Business in Huatai Securities,F832.51
  14. Classification of Customer Value Based on Rough Sets - Neural Network,F224
  15. The Improved Solution Research of Yiyang Tobacco Company’s CRM System,F426.8
  16. The Research of Customer Relationship Management for China Changjiang Bunker (Sinopec) Co., Ltd Yueyang Branch,F426.22
  17. A Case Study on the Customer Relationship Management of South China Network Power,F426.2
  18. Research & Application of Information Center of Customer Service in Changfeng Motor Company,F274
  19. The CRM Applications at Business Schools,F274
  20. The Design and Implementation of 4S Stores Management System Based on. NET Technology,TP311.52
  21. Shanghai GM Customer Management Information System Design and Implementation,TP311.52

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > General issues > Theories, methods > Algorithm Theory
© 2012 www.DissertationTopic.Net  Mobile