基于客户行为分析与企业销售预测的客户管理决策模型与算法研究-软件工程专业论文.docx

第 第 PAGE IV 页 第 第 III 页 统的客户行为分析进行改良，还要结合企业销售预测一起进行分析。通 过对二者的辨证分析，剖析出客户行为状态与企业预期业绩的动态变化 关系，并通过明确其相互关系对客户行为状态进行调整，使得基于客户 行为方面的决策更加准确，从而可以挖掘出企业的最佳客户决策策略， 以满足企业利益最大化。 本文首先分别对客户行为分析与企业销售预测做了综述，分析和比 较市面上用于二者的主要数学模型与算法，并在此基础上确定出相对比 较适合的模型与算法；其次，分别对所选用的模型与算法进行改进：客 户行为分析模型采用 Dirichlet-Multinomial 模型来估计转移概率，并 运用马尔科夫链蒙特卡洛模拟方法(MCMC)中的吉布斯(Gibbs)抽样算法 来优化该模型的参数，而企业销售预测模型通过采用加权思想对马尔可 夫链状预测模型的各阶状态转移概率矩阵进行加权求和，同时利用遗传 算法理论对建立的预测模型的状态转移概率矩阵的权重进行重新优化分 配；然后再结合实际的业务需求与相关的建模软件，对所提出的新模型 与算法进行检验；最后，重点分析和研究了客户行为状态与企业预期业 绩的动态变化关系，并在其二者之间进行相互佐证，进一步明确其相互 关系。 关键词：客户行为分析，CRM，RFM，马尔可夫链，状态转移矩阵 DECISION-MAKING MODEL AND ALGORITHM BASED ON CUSTOMER BEHAVIOR ANALYSIS AND SALES FORECASTS ABSTRACT With the advances in theory and technology， and the steady needs in enterprise market development, customer assets as an important intangible asset of enterprises, and its importance has received extensive attention as a key measure of corporate market value of one of the elements. Therefore, more customer relationship management software vendors attention the customer buying behavior. The basis work of customer management is customer analysis in modern enterprise. Not only help to adjust business strategies, but also enhance the competitiveness of enterprises. Eased the business operational risks in the unknown market environment. But the most scholars focused on customer segmentation, customer value analysis in the customer analysis, and the lack of research on customer behavior, or simply to use historical data to traditional regression analysis, the results is relatively crude, and cannot provide an effective decision-making for corporate. Even if some scholars based on behavior analysis, the status of behavior analysis is relatively simple, and lacked flexibility. But with the world economy development, and the preferences of customer have gradually become more diversified, and each customers characteristics and differences in buying behavior has become increasingly , the traditional customer co