A Systematic Literature Review on Applying CRISP-DM Process Model

TLDR

CRISP‑DM is the de‑facto standard process model for data‑mining projects across industries. The study systematically reviews recent CRISP‑DM applications to identify best practices, support phases for analysts, and provide a template for structuring and releasing such studies. The review surveys recent literature, summarizing research focus, current methodologies, best practices, and gaps across the six CRISP‑DM phases. CRISP‑DM remains the standard, yet most studies omit a deployment phase, posing a challenge.

Abstract

CRISP-DM is the de-facto standard and an industry-independent process model for applying data mining projects. Twenty years after its release in 2000, we would like to provide a systematic literature review of recent studies published in IEEE, ScienceDirect and ACM about data mining use cases applying CRISP-DM. We give an overview of the research focus, current methodologies, best practices and possible gaps in conducting the six phases of CRISP-DM. The main findings are that CRISP-DM is still a de-factor standard in data mining, but there are challenges since the most studies do not foresee a deployment phase. The contribution of our paper is to identify best practices and process phases in which data mining analysts can be better supported. Further contribution is a template for structuring and releasing CRISP-DM studies.