Abstract: The evaluation of associated rules is an important question in the process of data mining. Good evaluating methods are favor of removing feigned and unvalued associated rules. Traditional evaluating methods are support and confidence which are both based on probabilistic computing. The limitations of those two methods are analyzed. It is indicated that lacking logic basis is the important question that the methods of support and confidence are both in face of. With the theory of general relativity, a new method of abstracting associating rules is presented, which is based on formulas of general logics. Experimental results show that the method can improve the quality of abstracting process effectively.