DATABOOK : a standardised framework for dynamic documentation of algorithm design during Data Science projects

This paper proposes a standard documentation framework for Data Science projects, called Databook. It is a result of five years of action-research on multiple projects in several sectors of activity in France, and of a confrontation of standard theoretical Data Science processes, such as CRISP_DM, w...

Full description

Saved in:
Bibliographic Details
Published in:IASSIST quarterly Vol. 45; no. 2
Main Author: Nesvijevskaia, Anna
Format: Journal Article
Language:English
Published: International Association for Social Science Information Service and Technology 26.09.2021
Subjects:
Algorithm Transparency
Artificial Intelligence
Computer Science
Data Science
Documentation
Humanities and Social Sciences
Library and information sciences
Machine Learning
Project Process
Reproducibility
Statistics
ISSN:0739-1137, 2331-4141
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper proposes a standard documentation framework for Data Science projects, called Databook. It is a result of five years of action-research on multiple projects in several sectors of activity in France, and of a confrontation of standard theoretical Data Science processes, such as CRISP_DM, with the reality of the field. As a vector for knowledge sharing and capitalisation, the Databook has been identified as one of the main facilitators of Human Data Mediation. Transformed into an operational prototype of simple and minimalist documentation, it has since been tested then on about a hundred Data Science projects, has proven its benefits for the internal and external efficiency of Data Science projects, and can be turned into a more ambitious standard framework for data patrimony valorisation and data quality governance.
ISSN:0739-1137
2331-4141
DOI:10.29173/iq989