Research On Application Of Customer Buying Behavior Based On Business Intelligence

In the background of economic globalization, with the deepening of the use ofinformation technology, business intelligence technology began to be introduced bymore and more business fields, especially in such as banking, telecoms, insurance,transportation, retail and other fields, business intelligence success to help enterpriseintegrating data and extract useful information and convert to knowledge, to helpenterprises to realize more effective management, to make more wise decision, to gaingreater efficiency.This paper use business intelligence system of data mining technology, combinedwith retail enterprise sample database, to average purchase amount, purchasefrequency and the guest subdivision variables as age, to the customer purchasebehavior building customer segments of the model, and for each type of client forfeature extraction for analysis, for retail enterprise in the knowledge of customers anddeveloping marketing strategy to provide certain reference value.Customer clustering analysis, K-means algorithm is a kind of efficient algorithm.But the weaknesses of the algorithm was also evident, clustering to specified numberin advance, randomized initial clustering center clustering results of big effect. Andthe organization SOM neural network algorithm which can identify the sample datafrom a less than the number of unit of output, obviously for input data with "cluster"effect, but also for the next K-means the implementation of the clustering algorithmdesignated the number and the initial clustering center. So, this paper puts forward theSOM neural network algorithm and the K-means algorithm combined with clusteringanalysis method, the whole clustering analysis is divided into two phases: the firststage using SOM neural network get cluster number and clustering center; The secondstage with the first stage of the output as K-means the input of the algorithm.Because of clustering analysis is no guidance of study, of the classification of theclass is unknown, so this paper data after clustering analysis, using decision treealgorithm classification, to extract the common features of each type of customer base, for retail enterprise decision makers to provide valuable basis for decision-making.