Decision-making Model And Algorithmbased Oncustomer Behavior Analysisand Sales Forecasts

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 customer's characteristics and differences in buying behavior has become increasingly , the traditional customer concept will no longer available. Based on this market environment, the traditional unitary state analysis of customer behavior cannot meet the needs of diverse customer behavior. Therefore, it is necessary to improve the traditional way. Help enterprise find out the real value customers and provide a more effective decision-making.At the same time almost all of the analysis of customer behavior were not corporate with sales forecasts, but the two kinds mode of analysis are not isolated, should be inseparable. Therefore, the paper not only focus on improved the traditional analysis of customer behavior, but also analyzed together with enterprise sales forecasts. Through the dialectic between the two analysis, find out the dynamic relations between behavior analysis and sales forecasts, and adjust the customer behavior by the clear correlations, making the decisions based on customer behavior more accurately, which can be tapped out of the best customer decision-making strategy to meet the interests of enterprises to maximize.The contents of this paper are organized as follows:Firstly, we summarize the technologies about customer behavior analysis and sales forecasting respectively, analysis and comparison the major mathematical models and algorithms of market, and on this basis, identify relatively suitable model and algorithm; Secondly, we improve the selected model and algorithm respectively:customer behavior analysis model using Dirichlet-Multinomial model to estimate the transition probabilities, and using Markov chain Monte Carlo simulation methods (MCMC) in the Gibbs (Gibbs) sampling algorithm to optimize the model parameters, and the sales forecasting model through using weighted sum in the process of Markov chain forecasting model for each state transition probability, while the use of genetic algorithm theory, the establishment of predictive models of the state transition probability matrix to re-optimize the weight distribution; And then combined with actual business requirements and associated modeling software, testing the new model and algorithm; Finally, focusing on analysis and research the dynamic changes relationship between customer behavior states and business performance, and further clarify their mutual relations.