A software system maintenance activity is typically performed under an environment of lacking knowledge about how to process it. This scarcity of knowledge may be caused by various factors, such as the large size and complexity of the systems, high staff turnover, poor documentation and long-term system maintenance. The study applies Apriori algorithm to extract information from software change logs. Unfortunately, the software change logs generate many rules. Because searches the suitable rule from many rules is difficult and important matter, especially. This study focuses on the software co-change dependency and proposes a classification model based on association mining, to deal with such kind of dependency. The model combines data mining technologies, the traditional decision-tree and neural learning capabilities, to handle the complicated and real cases, and then improve the rule searching efficiency and the matching accuracy.