Privacy-preserving federated learning for interpretable student at-risk prediction across schools
Mahdee Jodayree, Arman Kavoosi Ghafi, Mostafa Atashafrouz, Mohammad Hossein Shafiabadi · 2026
Read Privacy-preserving federated learning for interpretable student at-risk prediction across schools by Mahdee Jodayree, Arman Kavoosi Ghafi, Mostafa Atashafrouz, Mohammad Hossein Shafiabadi (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13500831.txt.
Description
Mahdee Jodayree leads a four-author team with a 2026 framework for predicting which students risk failing, without pooling their records in one place. Schools increasingly want early-warning models, yet centralizing sensitive student data raises privacy and ownership concerns. Their system, called FL-AtRisk-DP-PBT, trains a shared logistic regression model across schools using federated averaging, so records stay local. Client-side differential privacy adds noise for protection, and a proof-before-train protocol logs each school's part so the process can be audited.
The team tested three setups on a main cohort of 14,003 students split into ten simulated schools, plus two public education datasets. The central model reached 99.14 percent accuracy, while the private federated version reached 98.00 percent, a small cost for the added protections. On the two public datasets the private version scored about 93 to 94 percent accuracy. Feature patterns stayed similar across setups, keeping the model interpretable.
Catalog record
| CREATOR | Mahdee Jodayree, Arman Kavoosi Ghafi, Mostafa Atashafrouz, Mohammad Hossein Shafiabadi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 202154 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |