Research On Attribute Reduction And Rule Acquisition Algorithm Based On Incomplete Decision Table

With the rapid development of high-tech, there are billions of people around the world in the use of mobile phones, computers and other electronic products every day, produced a large number of huge amounts of data. Big data from the Internet rapidly spread to other areas, from all walks of life have been data to the penetration, for the effective treatment for huge amounts of data to become one of the hot spot of current research. In 2015, during the two sessions of China, premier li signed concerned under the state council to promote "Internet+" guidance of action, to the effective combination between Internet and traditional industries to provide the policy basis. The development advantage of data mining is more and more prominent, occupied the world of the Internet.The so-called data mining is the data in the database, the use of some data processing methods, find out the intrinsic relationship between information process. Data mining includes mathematical statistics, pattern recognition, neural network and other technologies, which are widely used in the field of telecommunications, aviation, government and science. Rough set is one of the methods of data mining, and it is a tool to deal with the inaccuracy and the default value. Compared to fuzzy sets, the most prominent feature of Rough Set is that it does not need any other prior information, or additional knowledge, which can directly process information and find out the hidden knowledge in the mass data and used in real life.In the process of data mining, attribute reduction can be regarded as a preprocessing step, and it plays an important role in rough set theory. Attribute reduction can be defined as the premise of ensuring that the knowledge base can effectively distinguish the object, which can not play the role of the decision. Therefore, after the operation, it can effectively reduce the attribute of the knowledge base, and it also ensures the integrity of the basic information of the object. After the elimination of redundant attributes in the knowledge base, the space occupied by knowledge is reduced to a large extent, and the cost is reduced effectively.In the research process, the rough set attribute reduction is divided into two types, one is contains all of the information, called complete decision table and another containing uncertain or vacancy information, named incomplete decision table. Complete decision table refers to all the object information in the knowledge system is accurate and comprehensive. Now it is relatively comprehensive and mature to explore the attribute reduction of complete knowledge system, and it also has a great guiding role in the practical application. But in real life due to data collection equipment inevitable error, or collect the influence of outside, resulting in knowledge system is the lack of information, original complete knowledge of the system's algorithm cannot apply to this case. The research on incomplete decision table is a key concern of scientists and scholars in the future. In addition, rule acquisition is a more practical field.Enterprises and various organizations as the growth of the information explosion, accumulated the massive amounts of data, but the data itself is not useful information, the data of the rules is to help enterprise to make effective decision information, and provide reliable basis for the enterprise. Therefore, the rules in the decision table are also practical.In this paper, the research results obtained in the rough set attribute reduction and rule acquisition are studied, and the following innovations are carried out on the basis of the combination of them:(1)In the incomplete information system, the introduction of the granularity of discernibility matrix, on the idea of the condition attributes in distinguishing between objects appear frequency, on the granularity of discernibility matrix based on construct the heuristic function of attribute frequency calculation. A new reduction algorithm is put forward, which is based on the granularity of knowledge, and the complexity of space and time of the algorithm is greatly reduced, Respectively down to O(K|C||U|)(K= max{| TC(xi)|,xi ? U? and O(|U|). Finally, through the simulation experiment, it is proved that the algorithm has higher time efficiency compared to the same kind of algorithm.(2)In this paper, a defect of LEM2 rule acquisition algorithm is introduced, which has some attribute value pairs to be repeatedly included in the set of candidate attributes, and then it is eliminated in the back of the algorithm.And the deficiency of the improved algorithm is introduced. This paper introduces generalized decision function, the program executes before delete redundant attribute pairs, that they can't get into the back of the series of operations, reduces the attribute-value pairs sample space, improve the efficiency of calculating the attribute-value pairs.