banking-intent-classifier / verification.json
willykeenan's picture
Publish reproducible BANKING77 model and agent-evaluation evidence
438f1e0 verified
Raw History Blame Contribute Delete
1.29 kB
{
"claimBoundary": "Re-export and reproducibility check of an exploratory, test-informed model; not a new independent experiment.",
"dataAudit": {
"ambiguousTestGroups": 1,
"ambiguousTrainGroups": 1,
"categories": 77,
"cleanUniqueTrainRows": 9971,
"normalizedTestGroups": 3076,
"normalizedTrainGroups": 9972,
"normalizedTrainTestOverlapGroups": 25,
"postFilterOverlapGroups": 0,
"primaryUniqueTrainDisjointTestRows": 3050,
"rawTestRows": 3080,
"rawTrainRows": 10003
},
"elapsedSeconds": 7.958,
"kind": "frozen-recipe-export-verification",
"metrics": {
"accuracy": 0.9114754098360656,
"expectedCalibrationError": 0.14667641646587834,
"logLoss": 0.4433595294017486,
"macroF1": 0.9118581227515896,
"top3Accuracy": 0.9744262295081967
},
"modelSha256": "5f325e3b01103164f080e92a0df4f233402523a12f6a02c85e91ff7026e7f1d4",
"numpy": "2.0.2",
"python": "3.12.12",
"referenceCases": 3050,
"referencePredictionAgreement": 3050,
"referenceTop3PpmAgreement": 3050,
"restoredProbabilitiesBitwiseEqual": true,
"scikitLearn": "1.6.1",
"sourceCommit": "54041045c82953cbbabac155b150a0f7b7ed5603",
"sourceRepository": "https://github.com/willykeenan/waggle-kea",
"status": "passed",
"untrustedSkopsTypes": []
}