Knowledge Discovery Of Database And Its Application

The need to extract information automatically from very large databases has grown significantly in recent years. In response to this growing need, the closely related fields of Knowledge Discovery in Databases (KDD) and Data-mining have developed processes and algorithms to intelligently extract interesting and useful information (i.e. knowledge) from huge amount of raw data. Such techniques are used in various application domains, ranging from processes of production to business management. This paper studied the manufacturing processes of a large national steel- iron production corporation, and developed an improved method of KDD for the identification of the rules and knowledge for their desulphurisation processes.The study firstly concentrated on various areas of the KDD algorithms, Genetic Algorithm and their practical applications. Upon detailed analysis, and because of their limitations, the concept of Generalised Genetic Algorithm was introduced with related biological study explained. The study was further enhanced by investigating into the various flows with experimentations. The Chaos Search Algorithm was also examined and extended to confirm the availiabilty of RBF width.The basic theory of RBF Neural network was then analyzed. In order to overcome the difficulty in determining the RBF center numbers and spread, an improve method basing on the oforesaid Generalized Genetic Algorithm was then developed. Simulation results using the approach presented by J.D.Schafer show that the improved RBF neural networks are efficient and applicable for system modeling.In order to optimize the centers of RBF more effectively, the input-output clustering method is introduced. Because of the fact that the spread of RBF affects the generalization of the neural networks directly where large spread leads to inaccuracy while on the contrary, small spread harms the generalization. The Chaos Search Algorithm was therefore adopted to optimize the spread of RBF. Simulation results indicate that the Chaos Search Algoritmm using for RBF neural networks are efficient and useable.An iron water desulfurisation static prediction model was constructed by the optimized RBF neural networks in order to confirm the aforementioned analogy. The input parameters included the weight of iron water, quantity of sulphur before and after the desulphurisation while the output parameter is the weight of desulphurisation agent.An offline simulation was conducted and the results show the following conclusions:1. The improved generalized genetic algorithm for RBF not only can be used to determine the center numbers of RBF, but also optimize the location of these centers, hence improve the accuracy of the model;2. The Chaos Search Algorithm can be used to find the suitable spread which guarantee the good generalization of RBF neural networks;3. The iron water desulphurisation static simulation model based on RBF neural netwoks can be successfully applied to the prediction of the quantity of desulphurisation agents in steel making process.4. Although various algorithms have been used in the simulation models and real practical applications, ways and means to improve the converging speed is still worth further investigating.