Upload trained reconciliation model artifacts

README.md

@@ -1,3 +1,14 @@
----
-license: apache-2.0
----
+# Reconciliation Break Classifier
+
+This Space hosts a trained **ML-based reconciliation break classifier**.
+
+## Artifacts
+- `best_model.joblib` – trained ML pipeline
+- `feature_schema.json` – input feature definition
+- `metrics.json` – evaluation metrics
+
+## Usage
+Load using `joblib.load()` and apply on canonical reconciliation features.
+
+## Domain
+Finance · Banking · Transaction Reconciliation · Fraud Detection

best_model.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15a1156f016317b3ccb2479e49bf8304601ec0b431a5079eec0ee22c0cdd24f4
+size 3242

feature_schema.json

@@ -0,0 +1,26 @@
+{
+  "feature_cols": [
+    "amt_diff",
+    "amt_ratio",
+    "amt_pct",
+    "currency_match",
+    "time_gap_hours",
+    "settlement_gap_days",
+    "is_weekend",
+    "is_night"
+  ],
+  "numeric_features": [
+    "amt_diff",
+    "amt_ratio",
+    "amt_pct",
+    "currency_match",
+    "time_gap_hours",
+    "settlement_gap_days",
+    "is_weekend",
+    "is_night"
+  ],
+  "categorical_features": [],
+  "target_col": "label",
+  "created_at": "2025-12-24 04:59:44.498105+00:00",
+  "best_model_name": "logreg"
+}

inference_helper.py

@@ -0,0 +1,18 @@
+import json
+import joblib
+import pandas as pd
+
+def load_model(model_dir):
+    model = joblib.load(f"{model_dir}/best_model.joblib")
+    with open(f"{model_dir}/feature_schema.json") as f:
+        schema = json.load(f)
+    return model, schema
+
+def predict(df: pd.DataFrame, model, schema):
+    X = df[schema['feature_cols']].copy()
+    proba = model.predict_proba(X)[:, 1]
+    pred = (proba >= 0.5).astype(int)
+    out = df.copy()
+    out['pred_label'] = pred
+    out['pred_probability'] = proba
+    return out

metrics.json

@@ -0,0 +1,18 @@
+{
+  "logreg": {
+    "roc_auc": 1.0,
+    "pr_auc": 1.0
+  },
+  "random_forest": {
+    "roc_auc": 1.0,
+    "pr_auc": 1.0
+  },
+  "gradient_boosting": {
+    "roc_auc": 1.0,
+    "pr_auc": 1.0
+  },
+  "mlp_nn": {
+    "roc_auc": 0.9999879737703119,
+    "pr_auc": 0.9996221454837869
+  }
+}