Model card
astris-retention-mlp-v1
From-scratch numpy multilayer-perceptron retention model (manual backprop + Adam) predicting P(retain >= 90 days). No external deep-learning dependency.
At a glance
| Kind | scoring neural |
| Version | 1.0.0 |
| Status | Active — contributing to scores |
| Ensemble weight | 0.25 |
| Trained at | Not trained yet |
| Trained on | n/a |
| Artifact | No artifact (heuristic or pre-training) |
Metrics
{
"metric": "auc",
"target": "retain_ge_90d",
"validation": "synthetic_seed",
"real_cohort": "pending",
"value_approx": 0.8
}Notes
AUC ~0.80 on internal synthetic validation; real-cohort benchmark pending. Voter in the retention ensemble — see api/ml/neural.py.