基于粗糙集的数据挖掘算法分析与应用-管理科学与工程专业论文.docxVIP

Abstract As the key link of knowledge discovery, data mining can extract the valuable information from the massive data. And it is a hot field in the current artificial intelligence and information science research. Based on rough set, the data mining is the process of excavating novel and useful knowledge from data by using the rough set theory and method. The decision tree is a method commonly used in data mining. It has a simple structure, works very fast and it is very easy to understand. However, in the process of application, the present decision tree algorithm has many disadvantages. In this paper, further analysis and research will be done on how to optimize the decision tree algorithm by combining rough set and decision tree. The main research work is as follows. In the section of rough set, the algorithm of discernibility matrix reduction will be discussed and the method of simplifying of it will be put forward. Compared with single element structure discernibility matrix, the reduction of decision table is easier to get by using the equivalence class structure discernibility matrix. The bigger amount of knowledge of the attribute shows its distinguish force is more strong. A reduction method of decision table based on attribute distinguish force will be put forward. The method combines discernibility matrix and the attribute distinguish force. Compared with the methods such as the positive region, it can search for an attribute reduction more easily. In the section of decision tree, combined with rough set, the optimized decision tree algorithm based on the distinguish between value will be presented. The method divides the attributes into two fractions according to the difference of the distinguish between value. It can get the candidate attributes of a higher distinguish between value a little bit quickly. The algorithm of decision tree will be optimized from two aspects——knowledge reduction and pruning. The post-pruning optimized algorithm of decision