Applying Rough Sets Theory Into Practical Application

This thesis is circle around the point that how to apply Rough set theory into practice application and how to solve the probably difficulties lying ahead. To achieve this point, we discussed Rough sets' fundamental theory, the basic model of rough sets application, information table completion and data discretion. Introducing and reviewing of various reduction methods are also include in this thesis. Finally, two successful examples are introduced and some rough sets software toolkits are also have been detailed discussed. Thus a concrete rough sets application theory frame has been established.Rough sets' basic theory are reviewed, and rough sets' theoretical research condition are generalized and the current rough sets applications are listed. The general practical model of rough sets are listed and this thesis have reviewed the information completion and continuous discretion, common reduction algorithm are also introduced and discussed.A financial crisis prediction model combining with Rough sets and Kernel Fisher Discrimination(KFD) Methods is introduced. Rough sets theory reduction, a effective feature extraction method is utilized as data pre-preprocessing part and KFD methods are used as knowledge acquiring methods. Finally, this thesis put this model into the analysis of Chinese cooperate financial data, and the result of experiment verified the efficiency of this model.Another example is utilizing rough sets decision rules into character recognition. The thesis use pixel matrix as decision table and decision rules as classification method. Finally this method is proved its efficiency by experiment.The thesis also introduced the common rough set software toolkit. Each of them is being reviewed with its advantages and disadvantages. I believe this part would be great useful for those people who are interested in Rough Sets.In general, Rough Sets theory and methods have been proved as an effective method in Intelligent Information Processing and deserved to be researched and be put into practical. However, it still exists some fallback, which include dynamitic data support, very large database processing and redundant decision rules.