Multi-Classifier Combination Base On Rough Set And The Application Research On Knowledge Discovery In Database

Knowledge Discovery in Database (KDD) and Data Mining (DM) are Machine Learning methods in essence. And classification is the basis of the common Machine Learning problem. With the development of science and technology and the expansion of research work, the single-classifier can't satisfy application requirements of researchers. In this paper, problems in multi-classifier combination, KDD and DM are discussed and the model and methods of KDD applying Rough Set Theory, combined with Knowledge Discovery in Knowledge Database (KDK), are analyzed. The research work of this paper is following: A model of KDD based on knowledge base is put forward. It is an efficient Machine Learning method and can run better. It can do segmenting or dividing with super databases and also can be extended to the distributed data mining. It can be fit for the dealing with increment-learning algorithm and heterotgeneous data source. The corresponding algorithm based on RS is provided. Experimental results show out the probability and validity of the presented algorithm. Some solutions to contradictions or conflicts of regular knowledge are given out. The present corresponding algorithm is improved and the knowledge precision is advanced. The deeper knowledge discovery in knowledge base is focused on. And it is different from KDD&K because the later mainly learns from the discovered knowledge and then direct the KDD process. RS theory is introduced into KDD based on knowledge. And the idea effects well. RS theory is better than common methods of KDD and DM in some aspects. For example, it doesn't need the transcendental information, is easy to operate and its algorithms are simple.