A process model for design-oriented machine learning research in information systems

Abstract

This paper proposes a process model for design-oriented machine learning (DS-ML) research in the area of information systems (IS). As DS-ML studies become more prevalent in addressing complex business and societal challenges, there is a need for a standardized framework to conduct, communicate, and evaluate such research. We integrate elements from the design science research (DSR) process model, action design research (ADR), and the Cross Industry Standard Process for Data Mining (CRISP-DM) to develop a comprehensive Machine Learning Process Model (MLPM) tailored for academic DS-ML studies. The MLPM outlines eight key phases, including: problem identification; objective formulation; data understanding; data preparation; design, development, and refinement; evaluation; reflection and learning; and communication. We discuss the unique aspects of each phase in the context of DS-ML research and highlight the iterative nature of the process. By providing this structured approach, we aim to enhance the rigor, transparency, and comparability of DS-ML studies in IS research. This model serves as a step towards establishing consistent standards for DS-ML research, facilitating its integration into mainstream IS literature, and unlocking new opportunities for innovation and impact in the field.