Research On Application Of A Naivebayes Algorithm Based On The Rough Set Approach

The data mining is a very active part and one of the core contents of the current artificial intelligence research field. The rough set theory is a math tool able to deal with inaccurate and uncertain problems, have a firm theoretical basis.The research in this thesis focuses on the naive Bayesian classification model based on the weighting according to the attribute significance in rough set theory. In light of the often unavailability of those hypotheses in practical problems, the thesis makes a deep study in the combined application of naive Bayesian classification model and rough set theory in the engineering practice on the two aspects of application and theory. The work mainly includes three aspects as follows:(1) Systematically analyze the deficiencies of naive Bayesian classification model and present the directions of key improvement according to the demand of practical problems.(2) With an eye to the naive Bayesian classification model, analyze the theories and algorithms in the rough set theory that can improve its deficiencies and then present a new classification mode on basis of the combination. The classification model is based on attribute reduction, by means of attributing significance weighting, and finally calculates the weight of each attribute in the classification. The experiment shows that compared with the naive Bayesian classification model, the said new model can efficiently improve the classification effect and accuracy with the time consumptions are almost the same.(3) On basis of studying and analyzing the medical records in the hospital, present an integrated solution for the auto classification of medical records and intelligent medical guiding service. Firstly, collate large amount of medical records in hospital and then extract the useful data to formulate decision tables; secondly, simplify the data by means of attribute reduction; thirdly, evaluate the importance over the rest attributes and then do the 2nd weighting upon the descriptors to highlight the key words in medical records and user explanation, and take the 2nd weighting value as the corrected parameter to finish the final classified counting; finally, through the object-oriented analysis and programming, develop a naive Bayesian classification model system based on the weighting of rough set, successfully utilize it for the classification of medical records and intelligent medical guiding system in the medical field.The experiment shows that compared with the naive Bayesian classification model, the model mentioned in this thesis is more accurate and is a relatively successful classification mode.