Method mining in experimental software engineering

In experimental software engineering, particularly in the domain of predicting or identifying some properties (quality, cost, maturity, etc.), many methods have been proposed. The effectiveness of these methods depends on the software domain, the development process and style, etc., so there is no unique adequate method. The necessity of searching some adequate methods is recognized, but only few search techniques have been proposed thus far. Zhimin proposed a method mining technique for predicting error-prone modules based on a knowledge base consisting of a lot of prediction results. Here, we improve Zhimin´s method and propose a general method mining model for experimental software engineering, which can search for some adequate methods on the basis of the performance knowledge base. We also show several applications and their mining results, for example, error-prone module prediction and traceability link recovery.