Comparing Data Science Project Management Methodologies via a Controlled Experiment

Abstract

Abstract

CRISP-DM Twenty Years Later: From Data Mining Processes to Data Science Trajectories

“…Standardised processes are not the same as methodologies [53], and many methodologies do not necessarily include guided processes, where one can follow a series of steps linearly. Two cases that are close to data science are quite illustrative.…”

Thoughts on Current and Future Research on Agile and Lean: Ensuring Relevance and Rigor

“…Agile can be used in many other project contexts beyond software development, such as for data science [31], healthcare [39], hardware development [13], supply chain strategies [10] or for information security management [17].…”

Evolution Paths for Knowledge Discovery and Data Mining Process Models

“…Interactions between different roles in the process are partially addressed in models that identify the involvement of various human resources (see "Relevance of Human Role" section). On the other side, there is a current lack of adoption of mature team process methodologies to conduct data analytics projects; there has been little focus on how a team could effectively work together to do data science, nor much discussion has dealt with the team-based challenges that might occur when a group of people are working on a data science project [15]. RAMSYS is probably the first model that considers challenges deriving from the collaborative nature of a project when people involved are geographically dispersed [34]; the introduction of an information vault (Fig.…”

“…Saltz et al [15] compared the performance of four different methodologies/process models by conducting experiments on university students. The metrics used for the comparison were based not only on task output (i.e.…”

“…A blind application of data science techniques can lead to discovering meaningless patterns, and this can be prevented from occurring by formally structuring Knowledge Discovery and Data Mining (KDDM) processes [14]. Although it is likely that typical process steps, such as business objectives identification, problem definition, stakeholder identification, data architecture selection, analytical modelling and results validation, will naturally emerge through a data analytics project, the lack of a well-defined process model makes projects error-prone and devoid of any structured learning mechanism for future initiatives [15]. KDDM process models guide data project stakeholders to stay focused on the end product and prioritize goals for data analysis.…”