|
Based on sales of a tire company, a solution to data warehouse and data mining system is presented. Firstly, according to the sales business we establish multi-dimension data model. Secondly, extract historic data from source database and store it in data warehouse. Finally, focusing on the study of scalable classification algorithm that tightly integrates the building of decision tree with database technology, a combined optimization decision tree algorithm based on SQL is proposed.In this thesis, we analyze the procedure of data warehouse design, which includes data modeling, data warehouse building and administrating, discuss multi-dimensional modeling in detail. When analyzing dimensions, in order to get better drill-down path for sales, we implement the attributes conclusion and analysis methods, eliminate the irrelevant and redundant attributes.Compared with traditional classification algorithms, the algorithm makes improvements in several aspects, such as attribute selection, scalability and pruning. This combined optimization algorithm is based on weighed attribute’s consistency, using SQL to realize the computation data pre-processing, attribute selection measure and store decision tree in relational database. And the FAVC set make use of the processing capacity of large database and easily realized. At the end of paper, based on customer sales data, decision tree was built and stored in database, and customers were classified efficiently. Analysis results show that algorithm is better in many aspects than ID3 algorithm which is widely used in many fields currently.
|