Case study · Research
EduBridge Analytica — analytics educators can actually run
A research project asking whether a system-supported educational-data-mining workflow can make action research practical for non-technical educators — published with Henry Chukwudi John in the Babcock University Journal of Education (2026).
Fig. 01 — The EduBridge Analytica prototype. (Full system walkthrough screens coming soon.)
02 — The problem
Educators have the data — the workflow assumes an engineer
Learning platforms generate enormous amounts of data, and educational data mining and learning analytics offer established techniques to analyse it. But educator-led action research rarely benefits: the tooling presumes a technical background, and the earliest stages — data preparation and feature selection — carry the most cognitive load with the least support. The question we set out to answer: can a system-supported workflow make feature selection efficient, consistent and usable without displacing the educator's judgement?
Users are non-technical educators — no code, no schema wrangling.
Support must augment professional judgement, never substitute for it.
High-dimensional educational data — profiles, interactions, assessments — under real cognitive load.
03 — Process
From literature review to a measured system
STEP 01 · Frame
An integrated EDM–LA framework
The work began as a 2023 literature review at African Leadership University and grew into an integrated framework that organises analytical features by perspective (student, course, content, instructor) and source (profiles, performance, interactions) — giving the fuzzy early stages a structure educators can navigate.
STEP 02 · Build
The EduBridge Analytica prototype
A web application that operationalises the framework: guided data upload with quality preview, mapping onto a standardised schema, supported feature selection, and analytical dashboards for performance, engagement and course insights. I designed the framework, the prototype logic and the platform.
STEP 03 · Test
Two educators, three real questions
A mixed-methods study compared manual and system-assisted feature selection on a synthetic post-secondary dataset across three action-research questions — measuring completion time, feature overlap, and interview feedback on usability and cognitive effort.
Fig. 02 — Engagement analysis in the prototype: trends an educator can tie back to a research question (from the published article).
04 — Outcome
7.6× faster — with judgement intact
Educators completed system-guided feature selection dramatically faster, and their selections stayed aligned with what they would have chosen by hand — the system supported, rather than substituted, their judgement. Published as “EduBridge Analytica: Bridging Theory and Practice in Educational Data Mining for Action Research” with Henry Chukwudi John (African Leadership University) in the Babcock University Journal of Education, Vol. 11, No. 1 (2026), pp. 223–239 — DOI: 10.5281/zenodo.20712838. My contribution: conceptual design, the integrated framework, prototype logic and platform, and the quantitative analysis.