Data Mining Method Research Based On Rough Set And Particle Swarm Optimization

The Data Mining technique is a combination of machine learning, database and statistical theory. Data Mining can seek potential, interesting and valuable information. Rough Set theory is a mathematical tool used for dealing with vagueness and uncertainty. It is based on the indiscernibility relation that describes indistinguishable objects, and it does not require any additional empirical information of data sets, then can analyze and process the non-accurate, non–integrity and incomplete data,provides a new approach of data mining. In recent years, it has been successfully applied in many areas. Particle swarm optimization is an artificial technology. It is simple, effectual, quick convergent speed.This thesis studies about the theories of Rough Set and preprocesses of Data Mining by combining the Rough Set and Particle Swarm Optimization together, and studies a few problems of the theory. The following are some main points discussed by the thesis:1) The thesis reviews the existing approaches to processing missing values and analyzes their advantages and disadvantages in incomplete information systems, and compares the existing attribute reduction optimization and its drawback. The thesis imports the improved binary Particle Swarm Optimization, and giving the attribute reduction algorithm of incomplete dicision-making table based Rough Set and Particle Swarm Optimization. According to dependence of decision attribute vs. condition attribute, particle adaptive function is proposed based on approximate classified precision and the approximate classified quality. The experiment result the attribute reduction algorithm can reduce fast and efficient.2) The thesis designs a compound Particle Swarm Optimization by structuring compound particle struct, classifies the mix-data including the continuous attribute data and the discrete attribute data. Then the continuous attribute data need not discrete preprocessing, just compound operation combining the two processes of the continuous attribute data discretization and mix-data classification, It guarantees the classification object completely and reduces the error. The results of experiments prove that the algorithm reduces time and has nice computing performance.3 ) By analyzing the concept of network information system security from multi-criterion factors, a heuristic rough set algorithm and improved particle swarm optimization is proposed to reduce the attributes and build weight set. In addition, an information system security model is established.The examples show the feasibility and validity, and the practical value.