Granular Ranking Algorithm Based on Rough Sets

Abstract

Compared with the rule form in traditional data mining techniques, expressing knowledge in the form of rank list can avoid many disadvantages, and may be applied for the investigation of targeted marketing widely, identifying potential market values of customers or products. Based on the rough sets theory in granular computing, this paper proposes a Granular Ranking Algorithm with the time complexity O(nm), gives the framework of algorithm and the concrete algorithm steps. The core of new algorithm is the construction of Granular Ranking Function r _G (x) , which guides instances in the testing dataset finish ranking. The ranked result has a strong readability. The new algorithm improves the computation efficiency further relative to existing algorithms, e.g. the Market Value Function. The experiment result shows that the computation accuracy of granular ranking algorithm approaches to the market value function. Meanwhile, the time consumption of the former one is much less than the latter.