Research And Application On The Incremental Bayesian Classifier And Incremental Rough Set Algorithm

Bayesian classifier and rough sets are important technologies in the data mining domain,widely used in various fields such as in the economic, financial, scientific observations and engineering and so on.In the paper,how to get the core was elaborated in knowledge representation system after the main principle and method of rough sets was introduced.Then the mining model based on static reduction and dynamic reduction was separately analyzed.In the end,an algorithm was discussed, which could quickly get dynamic core weighted by the probability of sub-sets.The theoretical basis of sub-sets probabolity weight was that sampling different number of sub-sets had differently effect factors to decision table.The analysis had proved the model feasibility.During the case of higher dimentional attridutes and huge data sets, Bayesian classifier's parameters training was easily disturbed by noise and the classifier's robustness was not good. In the meanwhile, the complexity of incremental Bayesian classifier's damage assessment function was too high. In order to solve this problem, an incremental Bayesian classifier learning algorithm based on dynamic reduction had been proposed. The algorithm firstly analized the original data sets and mined the most favorable condition attributes for the classifier, named the minimum condition attributes. And the incremental Bayesian classifier could be trained and learned only by these most favorable condition attributes. Because of reduceing the number of condition attributes, that could reduce the complexity of classifier's damage assessment function and enhance the classifier's incremental learning efficiency. A series of experiments had been done to prove the algorithm is practicable.At the last, an application of the incremental Bayesian classifier algorithm and incremental mining algorithm based on rough sets was presented, in the alumina rotary kiln. At the same time, a model of incremental Bayesian classifier based on dynamic reductione was proposed in thermal data prediction. Futhermore, running test of the model had received good performance in field.