Research On Method Of Attribute Weight Based On Rough Sets Theory

Rough set theory is a tool, it is interpreted as a mathematical tool, which can process inaccuracy and incomplete information. Rough set theory is first proposed by Palawk, a Polish scientist. Since the rough set theory uses data instead of priori knowledge to process information, it embodies the idea of "speak with data". As a mathematical tool to process information, rough set theory has been widely used in some fields such as artificial intelligence and cognitive science. And attribute weight is one of this cores of the rough set theory. Attribute weight is important for the evaluation and decision process though attribute weight not only can directly affect the decision results, it also can directly affect the final judgment. Attribute weight reflects the position and function of the various factors in the decision evaluation process.Currently, many scholars have proposed methods of attribute weight based on rough sets theory. Among them, some use complete information system as their studying object, others use incomplete information system. So based on the analysis of many present methods of attribute weight based on rough sets theory, this paper’s research work is as follows:Firstly this paper studies the weight determining methods, based on positive region and conditional information entropy. According to the definition of the positive region and the conditional information entropy, it designed a new weight calculating method, which combined subjective experience with objective one, but still the reduction is too high, so modified attribute importance is then being put forward to solve this problem. It designed a method of rough set based on conditional information entropy’s weight determining in the complete information system. This method combines the subjective experience with objective one by using the definition of conditional information entropy, and overcomes the problem of too high reduction, which leads to non-redundant attributes being reduced to redundant attributes, therefore gets higher efficiency.This paper studies method of ascertaining attribute weight based on rough sets conditional information entropy, firstly seek the condition attribute the importance degree, based on it, and then use the attribute importance for the attribute weights. But the method of ascertaining attribute weight based on rough sets conditional information entropy still can’t guarantee that the calculated attribute weights is not0. So we propose an improved method, which use each conditional attribute’s importance and the number of value determined by it to get a new definition of the importance and weight of attribute. This method gets over the previously mentioned problem and is proved more efficient.Weight determining method has been a hot research topic in the filed of decision-making and evaluation. Compared to decision table, the importance of the conditional attribute is different, although some objects contain relatively more attributes which have unknown values, the importance of them are so low that we can consider them indistinguishable, while others have smaller number but all are very important, we think they are distinguishable. So, we should consider the number of attribute as well as its importance. This paper also introduces the concept of distinguish degree to determine the attribute weight in the incomplete information system, and based on this new concept we define a method to measure attributes’ importance, then design a new method to compute attributes’weight.Currently, research on the weight determining methods of incomplete information system are still very little, most of the used methods come from the improved methods used in complete information system, In order to get further improvement and obtain more reasonable weight, this paper proposes a new weight determining method. This method mainly uses the definition of conditional entropy of incomplete information system, combining the overall importance of conditional attributes with individual importance in system and giving new definition to attribute importance and the fast formula to calculate it. On this base we give out a new formula for calculating attribute weight. Since this method makes the calculated attribute weight more reasonable by combining the overall importance with individual one, the efficiency has been obviously improved.