Most SE data mining studies rely on well-known, publicly available tools such as association-rule mining and clustering. Such black-box reuse of mining tools may compromise the requirements unique to SE by fitting them to the tools’ undesirable features.