No More False Alert: Contrastive Learning for Predicting Health Deterioration from Imbalanced Care Records
Haru Kaneko, Sozo Inoue · 2026
Read No More False Alert: Contrastive Learning for Predicting Health Deterioration from Imbalanced Care Records by Haru Kaneko, Sozo Inoue (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/pmc-13259127.txt.
Description
Haru Kaneko and Sozo Inoue cut false health alerts in care homes with a contrastive learning method in a 2026 paper. Daily care records plus weather data feed a model that warns of next day deterioration in older residents. Rare events make ordinary models cry wolf, which exhausts staff.
Their training loss pulls truly worsening cases together while pushing false alarms away from them. Separation is shaped directly among cases the model flags as risky. Precision more than tripled against a resampling baseline in tests. Overall detection quality fell only slightly in exchange. Care managers gain an alert design that respects staff attention as a scarce resource.
Catalog record
| CREATOR | Haru Kaneko, Sozo Inoue |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |