Based On Conditional Granularity Entropy For Dynamic Attribute Reduction In The Incomplete Information System

Because of the big data era, we are faced with a difficult task:How to excavate potential, valubale information from the huge, messy, noisy, incomplete and fuzzy data? That is data mining. The classification is an important research branch in the data mining domain. By now many kinds of classication techques have been proposed and investigated. Compared with other classification methods, rough set theory is an useful mathematical tool for dealing with fuzzy and uncertain information without any prior knowledge beyond the data sets. So rough set has drawn much attention of many researchers.Attribute reduction is one of the kernel components in the rough set theory. At present, many attribute reduction algorithms of rough set under complete information systems have been proposed. But in real world application, there exit many such information systems that their partial conditional attribute values are unkonwn or missing and the data in the information systems are increasing dynamically, the classical rough set theory can’t cope with the kinds of incomplete information systems. Although many scholars have been studied them, in view of that dynamic attribute reduction algorithms in the incomplete information systems have not yet been sufficiently discussed so far. So it is of great urgency and necessary to research on dynamic attribute reduction under incomplete information systems.Based on the above, the research work of this paper are as follows:(1) Summarizing the current attribute reduction algorithms of rough set under incomplete information systems based on the viewpoint of information theory and algebraic theory.(2) From the point of view of information theory, we proposed a new form of conditional granularity entropy and attribute importance based on tolerance relation in the incomplete information system. And getting heuristic information from attribute importance, we presented the attribute reduction algorithm under the incomplete information system.(3) Further, we discussed the changes of conditional granularity entropy under three cases:only one object increases, only one object exits, one object increases and one object out at the same time, and updated the reduction algorithm correspondingly. Finally, the validity and the feasibility of the algorithm is demonstrated by applying the data of ship fuel consumption.