Create test.py

test.py

@@ -0,0 +1,111 @@
+import joblib
+import pandas as pd
+
+# === Load pretrained components ===
+tfidf = joblib.load("models/tfidf_vectorizer.pkl")
+models = joblib.load("models/logreg_model.pkl")
+encoders = joblib.load("models/label_encoders.pkl")
+
+# === Sample structured input (as in your app.py logic) ===
+sample_input = {
+    "Transaction_Id": "T001",
+    "Origin": "USA",
+    "Designation": "CEO",
+    "Keywords": "sanctioned entity",
+    "Name": "John Doe",
+    "SWIFT_Tag": "TAG001",
+    "Currency": "USD",
+    "Entity": "ABC Corp",
+    "Message": "Wire transfer from XYZ",
+    "City": "New York",
+    "Country": "USA",
+    "State": "NY",
+    "Hit_Type": "Manual",
+    "Record_Matching_String": "ABC",
+    "WatchList_Match_String": "XYZ",
+    "Payment_Sender_Name": "Jane Smith",
+    "Payment_Reciever_Name": "Robert",
+    "Swift_Message_Type": "103",
+    "Text_Sanction_Data": "Possible match with OFAC list",
+    "Matched_Sanctioned_Entity": "XYZ Entity",
+    "Red_Flag_Reason": "High Risk",
+    "Risk_Level": "High",
+    "Risk_Score": 85.5,
+    "Risk_Score_Description": "Above threshold",
+    "CDD_Level": "Enhanced",
+    "PEP_Status": "Yes",
+    "Sanction_Description": "Description text",
+    "Checker_Notes": "Some note",
+    "Sanction_Context": "Transaction linked with blacklisted entity",
+    "Maker_Action": "Hold",
+    "Customer_Type": "Individual",
+    "Industry": "Finance",
+    "Transaction_Type": "Credit",
+    "Transaction_Channel": "SWIFT",
+    "Geographic_Origin": "North America",
+    "Geographic_Destination": "Europe",
+    "Risk_Category": "Sanctions",
+    "Risk_Drivers": "PEP, Country",
+    "Alert_Status": "Open",
+    "Investigation_Outcome": "Pending",
+    "Source_Of_Funds": "Salary",
+    "Purpose_Of_Transaction": "Investment",
+    "Beneficial_Owner": "John Doe"
+}
+
+# === Combine into text format (same as in FastAPI logic) ===
+text_input = f"""
+Transaction ID: {sample_input['Transaction_Id']}
+Origin: {sample_input['Origin']}
+Designation: {sample_input['Designation']}
+Keywords: {sample_input['Keywords']}
+Name: {sample_input['Name']}
+SWIFT Tag: {sample_input['SWIFT_Tag']}
+Currency: {sample_input['Currency']}
+Entity: {sample_input['Entity']}
+Message: {sample_input['Message']}
+City: {sample_input['City']}
+Country: {sample_input['Country']}
+State: {sample_input['State']}
+Hit Type: {sample_input['Hit_Type']}
+Record Matching String: {sample_input['Record_Matching_String']}
+WatchList Match String: {sample_input['WatchList_Match_String']}
+Payment Sender: {sample_input['Payment_Sender_Name']}
+Payment Receiver: {sample_input['Payment_Reciever_Name']}
+Swift Message Type: {sample_input['Swift_Message_Type']}
+Text Sanction Data: {sample_input['Text_Sanction_Data']}
+Matched Sanctioned Entity: {sample_input['Matched_Sanctioned_Entity']}
+Red Flag Reason: {sample_input['Red_Flag_Reason']}
+Risk Level: {sample_input['Risk_Level']}
+Risk Score: {sample_input['Risk_Score']}
+CDD Level: {sample_input['CDD_Level']}
+PEP Status: {sample_input['PEP_Status']}
+Sanction Description: {sample_input['Sanction_Description']}
+Checker Notes: {sample_input['Checker_Notes']}
+Sanction Context: {sample_input['Sanction_Context']}
+Maker Action: {sample_input['Maker_Action']}
+Customer Type: {sample_input['Customer_Type']}
+Industry: {sample_input['Industry']}
+Transaction Type: {sample_input['Transaction_Type']}
+Transaction Channel: {sample_input['Transaction_Channel']}
+Geographic Origin: {sample_input['Geographic_Origin']}
+Geographic Destination: {sample_input['Geographic_Destination']}
+Risk Category: {sample_input['Risk_Category']}
+Risk Drivers: {sample_input['Risk_Drivers']}
+Alert Status: {sample_input['Alert_Status']}
+Investigation Outcome: {sample_input['Investigation_Outcome']}
+Source of Funds: {sample_input['Source_Of_Funds']}
+Purpose of Transaction: {sample_input['Purpose_Of_Transaction']}
+Beneficial Owner: {sample_input['Beneficial_Owner']}
+"""
+
+# === Vectorize ===
+X_tfidf = tfidf.transform([text_input])
+
+# === Predict ===
+print(" Predictions:")
+for label, model in models.items():
+    proba = model.predict_proba(X_tfidf)[0]
+    pred_idx = proba.argmax()
+    decoded = encoders[label].inverse_transform([pred_idx])[0]
+    print(f"{label}: {decoded}")