Weighting Of Data Mining Algorithms And Their Implementations In Business Intelligence Applications

With the development of computer technology, enterprises have entered the information age. Enterprises have accumulated a large number of daily data, which lead to the phenomenon of huge amount of data with poor knowledge. Data mining solves this problem, which can mine information from data, and can help the manager of enterprises to make decisions . Business intelligence is the specific application of data mining, so this dissertation researches business intelligence. Business intelligence contains data warehouse, ETL (Extraction Transformation Loading), OLAP (On-Line Transaction Processing), data mining .Since in practical application, various factors have different levels of important to result, so first this dissertation use method of serial replacement to determine the degree of importance of each factor. Then this dissertation use analytic hierarchy process to assign weight of each factor, data mining program realized on this base. This dissertation presents the algorithm to compute weight ,this method can ruduce rules number and foucs on mining user interest rules. Building FP-tree which use this dissertation presents method to calculate support , this algorithm can find both positive and negative association rules. Weighted least square method model dependent variable and independent variables to realized prediction function. This dissertation present use weights which determine by method of serial replacement and analytic hierarchy process replace weighted least square method's weights. Experiments show this method of determining the weights has a good degree of fitting. Naive bayes algorithm requires independent attributes, so this dissertation improves weighted mixed naive bayes algorithm,this dissertation use method which choosed to compute weight . weighted mixed naive bayes algorithm divides properties to independent subsets. Weighted TAN is used in subsets, naive bayes is used between these subsets. weighted mixed naive bayes algorithm overcomes the shortcomings of naive bayes and has good accuracy.In this dissertation do experiments to test the accuracy and effectiveness of above algorithm. Experimental results show that the above weighting algorithm not only reflect the impact of weight, but also can be use in data mining effectively. Majority of enterprises use ERP software to manage daily work. ERP software's databases contain large amount of data which can be provided to business intelligence. This dissertation presents an business intelligence and ERP integrated architecture. This dissertation realizes business intelligence function by secondary development of ERP software. This dissertation realizes the entire process of business intelligence which include following steps. This dissertation uses data smoothing, data standardization and range normalization to realize data extraction. This dissertation realize data mining program. This dissertation implement business intelligence module in ERP software.In this dissertation , experiments are done to verify the accuracy and validity of the alogorithm. This paper implement three data mining algorithms to verify bussiness inteslligence module can effectively run dan find correct rules.