Data mining for software engineering and humans in the loop

Abstract

The field of data mining for software engineering has been growing over the last decade. This field is concerned with the use of data mining to provide useful insights into how to improve software engineering processes and software itself, supporting decision-making. For that, data produced by software engineering processes and products during and after software development are used. Despite promising results, there is frequently a lack of discussion on the role of software engineering practitioners amidst the data mining approaches. This makes adoption of data mining by software engineering practitioners difficult. Moreover, the fact that experts’ knowledge is frequently ignored by data mining approaches, together with the lack of transparency of such approaches, can hinder the acceptability of data mining by software engineering practitioners. To overcome these problems, this position paper provides a discussion of the role of software engineering experts when adopting data mining approaches. It also argues that this role can be extended to increase experts’ involvement in the process of building data mining models. We believe that such extended involvement is not only likely to increase software engineers’ acceptability of the resulting models, but also improve the models themselves. We also provide some recommendations aimed at increasing the success of experts involvement and model acceptability.