Search

This paper proposes a standard documentary framework called Databook for Data Science projects. This proposal is the result of five years of action-research on multiple projects in several sectors of activity in France and a confrontation of standard theoretical processes of Data Science, such as CRISP_DM, with the reality of the field. The minimalist and flexible structure of the Databook prototype, described and illustrated in this paper, has revealed its operationality on more than a hundred projects and has been recognised by various stakeholders as an excellent facilitator of Human Data Mediation, especially for multi-skilled projects. Beyond its proven benefits for project efficiency, this framework, conceived as a frontier object, can be applied more broadly to data project portfolio management and data value, governance and quality. By surpassing the computational aspect of the models, the Databook is an answer to the issues of interpretability and auditability of algorithms. [ABSTRACT FROM AUTHOR]

During this digital era, France, like many other countries, is undergoing a transition from paper-based assessments to digital assessments in education. There is a rising interest in technology-enhanced items which offer innovative ways to assess traditional competencies, as well as addressing problem solving skills, specifically in mathematics. The rich log data captured by these items allows insight into how students approach the problem and their process strategies. Educational data mining is an emerging discipline developing methods suited for exploring the unique and increasingly large-scale data that come from such settings. Data-driven methods can be helpful when trying to make sense of process data. However, studies have shown that didactically meaningful findings are most likely generated when data mining techniques are guided by theoretical principles on subjects' skills. In this study, theoretical didactical grounding has been essential for developing and describing interactive mathematical tasks as well as defining and identifying strategic behaviors from the log data. Interactive instruments from France's national large-scale assessment in mathematics have been pilot tested in May 2017. Feature engineering and classical machine learning analysis were then applied to the process data of one specific technology-enhanced item. Supervised learning was implemented to determine the model's predictive power of students' achievement and estimate the weight of the variables in the prediction. Unsupervised learning aimed at clustering the samples. The obtained clusters are interpreted by the mean values of the important features. Both the analytical model and the clusters enable us to identify among students two conceptual approaches that can be interpreted in theoretically meaningful ways. If there are limitations to relying on log data analysis in order to determine learning profiles, one of them is the fact that this information remains partial when it comes to describing the complete cognitive activity at play, the potential of technology-enriched problem solving situations in large-scale assessments is nevertheless obvious. The type of findings this study produced is actionable from teachers' perspective in order to address students' specific needs. [ABSTRACT FROM AUTHOR]