Downloads · 30 days
5
17% of all-time downloads
Academica/AcademicaMed-ORACLE-ClinicalTrials-SuccessProb-v1
AcademicaMed-ORACLE-ClinicalTrials-SuccessProb-v1 is a tabular classification model from Academica. Use it for the tabular classification task on the model card, and read the license before you ship it in a product. The card lists the license as cc-by-nc-nd-4.0.
ORACLE is the synthesis classifier in the AcademicaMed clinical-trial model family.
Downloads · 30 days
5
17% of all-time downloads
All-time downloads
30
Public
Repo size
—
Likes
0
Public
Click a slice to open those files.
.md1.8 KB · 40%
From the Hugging Face model README
ORACLE is the synthesis classifier in the AcademicaMed clinical-trial model family.
The classifier reads concatenated PACT, MOAt, and FATE embeddings. It returns a calibrated success probability and a risk tier.
| Property | Value |
|---|---|
| Model type | Gradient-boosted ensemble with Platt calibration |
| Input | 768 dimensions: PACT 256, MOAt 256, FATE 256 |
| Output | Calibrated probability and risk tier |
| Training set | 12,692 completed trials |
| Validation set | 2,720 trials |
| Test set | 2,720 trials |
| Positive rate | 58.5% |
| Split | ROC-AUC | Average precision | Brier score |
|---|---|---|---|
| Train | 0.791 | 0.829 | 0.185 |
| Validation | 0.736 | 0.776 | 0.202 |
| Test | 0.734 | 0.782 | 0.205 |
Trial text -> PACT, MOAt, and FATE embeddings -> 768-dimensional vector -> classifier -> calibrated probability
ORACLE reflects historical records and labels. It does not provide medical advice or establish a trial's future outcome.
Performance measurements require independent validation before clinical, regulatory, investment, or operational use.
This repository contains the original published model card and configuration scope. The source repository did not publish executable classifier files.