Research On Discretization Of Attributes Based On Granular Computing And Rough Set

With the rapidly increasing amount of information in the system information of the database, data mining and other relevant fields in today’s society have become a research hot topic. However, in the real world the data collected directly tends to be continuous, what’s more, most of the data mining technology and machine learning algorithm can only deal with discrete attributes. So as the important preprocessing technology in the extraction of rules of data mining and machine learning, discreting Continuous attribute directly relate to the final effects of data mining or machine learning. Since the coming of the rough set and granular computing theory, many researchers have reference to and improved the existing discretization algorithm so as to solve the related problems in the field, and get good effect.Firstly, this paper introduces the particle, Granular layer and grain structure of granular computing problem, thus drawing the granular computing can efficiently solve some complex practical problems. What’s more it explains the basis of rough set theory and the standard of discretization and several classic proposed discretization algorithm in detail.Secondly, After studying the attribute discretization algorithm based on granular computing, this paper on the choice of the Section granular brings in a candidate’s breakpoint algorithm applying the orderly sequence. In addition,it also takes the size of the discrete intervals and the number of category in every interval into consideration improving the learning accuracy of the algorithm.To ensure minimum information loss, it uses level of consistency in rough sets theory in the control of the whole process of the algorithm.Finally, on the basis of the greedy algorithm, this paper brings in the concept of the ranks of importance, in the binary discernible matrix if calculate the importance of each candidate breakpoint, you can judge the importance of candidate breakpoint on each line and column, so as to extract the best candidates for breakpoints. What’s more, rows of the matrix bit operate with the breakpoint binary bit, saving more operation time.