基于判别向量的不一致决策表规则提取算法

Distinguish Vector Based Rule Acquisition Algorithm for Inconsistent Decision Table

规则提取广泛应用于机器学习和数据挖掘中,是一种获得隐含知识的理论方法。针对不一致决策表,从多粒度角度出发,将不一致决策表转换为一致决策表,然后定义判别向量,在由粗到细的粒度空间下分别对决策表进行分析,提出了一种新的不一致决策表规则提取算法,并通过理论证明和实例分析说明了新算法的正确性,最后由UCI数据集的测试对比了现有的一些不一致决策表规则提取算法的性能,测试结果显示了新算法的有效性和快速性。

Decision rule acquisition is an important aspect in data mining and machine learning. It is a theoretical method to obtain implicit knowledge. A new method for rule acquisition in inconsistent decision table is proposed, firstly inconsistent decision table is transformed into consistent table, then, from the perspective of multi-granulation, rules are extracted from coarse-to-fine granularity space by the definition of discrimination vectors. The correctness of the new algorithm is illustrated by examples and proved by theoretical proof. Finally, UCI data set are applied to test and compare the efficiency of the proposed algorithm with existing ones. The test results show its efficiency and rapidity.