Research And Application Of Mining Inter-transaction Quantitative Association Rules

The World has entered the era of information, there are a great deal of data in all sorts of fields. It is very urgent for people to transform them into useful information and knowledge, and apply them in business administration, production control and forecasting, etc. As a kind of technology for extracting information from large quantity of data, Data mining and knowledge discovery has become a significant research problem which has important theoretical and practical values, and attracts widely attention in international academe. Upon investigating into the research achievements and disadvantages of existing in knowledge discovery in database and in data mining, this thesis advanced the research and application of mining the inter-transactional quantitative association rules.First of all we introduced the up-to-date developing trends of domestic and abroad regarding knowledge and data mining, basic concepts and general processing of data mining. Through the analysis of inter-transactional association rules, we find it can only apply to categorical attributes, thus its application is limited, then we proposed the concept of inter-transactional quantitative association rules and corresponding mining algorithm, thus united categorical and quantitative attributes and greatly expanded their application range. Afterward we introduce partial completeness measure and box-dividing technology to dynamically discretize the attributes. They are capable of efficiently reduce the search space of algorithm in the precondition of minimum lost information. For the large quantity of result rules, we adopt the R-interest measure to cut those useless ones. Since data mining is essentially a kind of data analysis technology, pure data mining perhaps will not produce user-expected results. Thereby we introduce multidimensional data-oriented OLAP technology, and closely integrate them into a user friendly, interactive data mining environment. Finally the theory is applied into the design of Wuhan city air quality forecasting system-Orpheus. The main research fruits are as following:1) The theory and methods of inter-transactional association rules are studied intensively, the concept of inter-transactional quantitative association rules are proposed through expanding original concepts.2) The algorithm of mining inter-transactional quantitative association rules is propose.3) The concept of data warehouse and OLAP technology are introduced, and the system structure of OLAM is built on OLAP and association rules mining algorithms., and the system is implemented in air quality forecasting system.4) The framework of forecasting system is designed based on association mining algorithms and OLAM structure, the core of which is the OLAM engine. It implemented a auto-feedback process from data modeling to forecasting results to adjust the model. Therefore it is of illuminative significance for data mining in forecasting field.