Research On Technologies Of OLAM And Multidimensional Visualization On Instances

With the permeation of computers, the development of large capacity storage technology and the wide application of data gaining equipment such as bar code, RFID, and so on, great amount of various kinds of data has been accumulated in the routine transaction processing and scientific research. The historic data contains important information which can be referred to for making decisions. Therefore, now people attach much attention to how to make full and effective use of the data. Data warehouse, online analysis and processing, and data mining technology provide effective way in solving the problems of data exploding and little knowledge.Visualization technology can realize interactive analysis of direct data input, results output and mining, and provide ways of data mining with the participation of people's perception, insight and perspicacity, as well as ways of visualized mining and the visualization of mining results. It is very easy to find out key variables from numerous and complicated variables with visualized mining. The visualization of mining results can provide users direct results easy to be understood finally. Therefore, it is necessary to improve the efficiency of data mining, the reliability of results, and the combination of visualization and data mining.How to apply technologies related to data warehouse and data mining to customer relation management in manufacturing is the research field cried for by manufacturing. It includes online analysis of data warehouse and data mining, constructing analyzed CRM system, design and application of effective mining algorithm, and OLAM visualization. In this thesis, some key technologies of actualizing CRM system in manufacturing are discussed in details. Customer analysis mining prototype system on the basis of OLAM has been developed after the research of CRM technology based on OLAM and its visualization.OLAM which is the combination of DM and OLAP was researched theoretically. The core technologies of analysis CRM were discussed. After transaction analysis of a certain manufacturing, the data warehouse of CRM was constructed. It makes basis of further analysis of customer data and mining selling rule of customer products, and scientific division of customers.Three indexes of customer division—customer lifetime value, customer loyalty and customer credit—were calculated with corresponding algorithms:(1) Analyzed on the content of customer lifetime value; Researched the significance and influence factor of customer lifetime value model based on allotting expense; Proposed the means of calculating customer transferring matrix based on Markov Chain according to non-back effect of customers' shopping transfer; exactly calculated the influence which customer transfer has on customer lifetime value, combined with the case.(2) Researched on related theories of customer loyalty; Set up the index system of customer loyalty degree; Proposed forecasting customer loyalty model calculated with fuzzy neural net, because of complicated loyalty calculation; To ensure the efficiency of model training, the author proposed that attribute importance theory in the information field be used to determine initial weights of fuzzy neural net. And the experiment proved that it is effective.(3) Researched the theory of fuzzy appraisement and applied it to property reputation appraisement of enterprises and individuals; Created 2 kinds of index systems calculating credit; Determined the weights of the indexes with optimized choosing method; Analyzed the index affiliation influencing customer Property reputation with fuzzy theory; Finally customer property credit was calculated with software.The CLV/CL/CC customer division based on customer lifetime value, customer loyalty and customer credit was researched emphatically. On the basis of calculating and forecasting the factors of indexes, the results of clustering were calculated with improved K-means2 and the clustering was treated as the Bayes with attribute power count algorithm which was the step before division forecasting. The advantages of the two algorithms are combined together, which can realize customer division effectively and improve the efficiency of distinguishing customer division.The mining and online analysis of customers' shopping baskets is the key research point now. OLAM technology based on relevant rules and its visualization in multi-dimension have been researched. The data of product sales was analyzed with relevant time sequence and was visualized. The rules of the products relevance were analyzed, which can provide evidence for supporting decisions such as promotion strategy. The transaction data such as product sales volume and order list was analyzed on line through multi-dimensions and multi-levels up-drills, down-drills , horizontal or longitudinal section.And the customer property factors were analyzed with visualized and comprehended ways.On the basis of research involved in this thesis, an online customer analysis mining system based on data warehouse was designed and realized with the theories combined with a certain company. The analysis module and algorithm model can be integrated into the system, which can realize customer lifetime value, customer loyalty and customer credit, divide the customers, analyze the product sales rules deeply, and realize the visualization of OLAM analysis. As a tools platform, it provides great support for further management and research, and effective way of supporting decisions for analysis people in these fields...