Optimizing Methods And Application Of Data Ming Error Systems Based On Error-eliminating Theory

As informationization and networking proceed, a large amount of data information needs to be processed and analyzed. During the information mining, what is of paramount importance belongs to the identified classification, prevention and elimination of error information with the help if the law and reason of error information. Once error information emerges, the reasons and methods need to be found out to eliminate them. Usually, with the alteration of time, space, mining conditions and techniques, and technique development, it is possible for a totally correct information system and decision to switch into an error one. To avoid that, existent errors ought to be eliminated and meanwhile, data-mining methods should be optimized.In1983, Chinese scholar Guo Kaizhong set up error-eliminating theory, thereby taking people’s knowledge and research into errors to a higher level. Error-eliminating theory applies mathematic methods to study errors in quantity and systematic methods to probe into the relations of errors. The achievements in error-eliminating research have helped to set up two error-eliminating theoretical tools:"ten five six three" theory and error system theory.Carrying on the previous scientific achievements, this dissertation attempts to probe into how to set up data-mining error system and its realizing methods on computers based on the theory of error set, error logic and error matrix. The purpose is to find out reasons and laws why errors exist in data-mining through the new algorithm models developed by the optimizing methods of data-mining error system.Based on the error set, error system, T transformation of error set and error matrix equation, data-mining systems are provided with the basis of error-eliminating theory as well as the relations and operations of its error subsystems in data-mining error system constructing. Besides, the definitions, classifications and types of data-mining associating error functions together with the function and operations of associating rules are studied. What’s more, this essay puts forward data-processing methods, T transformation and its types of the error data set after expounding error data processing methods.In addition, with the consideration of the classifying characteristics of data bank, methods of error-eliminating data mining, error-eliminating clustering data-mining are obtained based on error eliminating methods. Also, this thesis puts forward the classification of error data set, the method to construct and abstract associating error function, time and space matrix clustering methods, error matrix clustering methods as well as the steps and laws to classify and cluster errors. Meanwhile, its functions in processing error data and optimizing data system are introduced.Moreover, based on the theory of error set, fuzzy error set and relative entropy, this dissertation sets up data-mining error system of dynamic information, studies the time-lag phenomenon, entropy changes together with optimization and worsening discerning. Their nature and characteristics are also discussed and meanwhile, the definition of the fuzzy error data set, multivariate error data set and multivariate error data of critical point are given.Lastly, after analyzing two cases, the paper introduces the error-eliminating classification and clustering methods of data-mining error systems into studying the examples and also constructs models of error-eliminating cluster with the realization of R compiling software on computers. At the end of the paper, examples are studied and analyzed in order to assist companies in making marketing and developing strategies.The creations of this paper residecin:(1) on the base of error-eliminating theory, the invistigation of error data set, fuzzy error data set, myltivariate data set and the error data set of critical point, the researches of error data-mining system of dynamic information together with the enriching and development of error-eliminating theory.(2) the introduction of error-eliminating classifying data-mining method models and error-eliminating clustering data-mining method models with the combination of data-mining theory, the enriching and development of data-mining methods, ehich laying a solid foundation for the future researches.