Create app.py

app.py

@@ -0,0 +1,90 @@
+import gradio as gr
+import torch
+from transformers import AlbertTokenizer, AlbertForSequenceClassification, BertTokenizer, BertForSequenceClassification
+import numpy as np
+import joblib
+import os
+
+# Load ALBERT model
+albert_model = AlbertForSequenceClassification.from_pretrained("Deepaksai1/albert-fraud-detector")
+albert_tokenizer = AlbertTokenizer.from_pretrained("Deepaksai1/albert-fraud-detector")
+albert_model.eval()
+
+# Load FinBERT model
+finbert_model = BertForSequenceClassification.from_pretrained("Deepaksai1/finbert-fraud-detector")
+finbert_tokenizer = BertTokenizer.from_pretrained("Deepaksai1/finbert-fraud-detector")
+finbert_model.eval()
+
+# Load CatBoost model
+from catboost import CatBoostClassifier
+catboost_model_path = "catboost_fraud_model.cbm"
+catboost_model = CatBoostClassifier()
+catboost_model.load_model(catboost_model_path)
+
+# CatBoost prediction (expects structured features, here we simulate with dummy value)
+def predict_with_catboost(text):
+    # Simulate with simple heuristic
+    amount = float([s for s in text.split(',') if 'Amount' in s][0].split(':')[1].strip())
+    prediction = catboost_model.predict([[amount]])[0]
+    proba = catboost_model.predict_proba([[amount]])[0][1]
+    return ("Fraud" if prediction == 1 else "Not Fraud"), float(proba)
+
+# ALBERT prediction
+def predict_with_albert(text):
+    inputs = albert_tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=64)
+    with torch.no_grad():
+        outputs = albert_model(**inputs)
+        probs = torch.nn.functional.softmax(outputs.logits, dim=1)
+        pred_class = torch.argmax(probs).item()
+        pred_prob = probs[0][1].item()
+    return ("Fraud" if pred_class == 1 else "Not Fraud"), float(pred_prob)
+
+# FinBERT prediction
+def predict_with_finbert(text):
+    inputs = finbert_tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=64)
+    with torch.no_grad():
+        outputs = finbert_model(**inputs)
+        probs = torch.nn.functional.softmax(outputs.logits, dim=1)
+        pred_class = torch.argmax(probs).item()
+        pred_prob = probs[0][1].item()
+    return ("Fraud" if pred_class == 1 else "Not Fraud"), float(pred_prob)
+
+# Main prediction selector
+def predict(text, model_name):
+    if model_name == "ALBERT":
+        return predict_with_albert(text)
+    elif model_name == "FinBERT":
+        return predict_with_finbert(text)
+    elif model_name == "CatBoost":
+        return predict_with_catboost(text)
+    else:
+        return "Unknown Model", 0.0
+
+# Example transactions
+examples = [
+    "Step: 305, Type: CASH_OUT, Amount: 2321633.57, Origin Balance: 2321633.57, Dest Balance: 0.0",
+    "Step: 6, Type: CASH_OUT, Amount: 13704.0, Origin Balance: 13704.0, Dest Balance: 3382.84",
+    "Step: 285, Type: TRANSFER, Amount: 125487.45, Origin Balance: 0.0, Dest Balance: 524556.64",
+    "Step: 352, Type: PAYMENT, Amount: 41263.42, Origin Balance: 0.0, Dest Balance: 0.0",
+    "Step: 372, Type: CASH_IN, Amount: 187503.32, Origin Balance: 76827.0, Dest Balance: 0.0"
+]
+
+# Gradio interface
+gui = gr.Interface(
+    fn=predict,
+    inputs=[
+        gr.Textbox(label="Enter Transaction Description"),
+        gr.Dropdown(choices=["ALBERT", "FinBERT", "CatBoost"], label="Select Model", value="ALBERT")
+    ],
+    outputs=[
+        gr.Label(label="Prediction"),
+        gr.Number(label="Fraud Probability")
+    ],
+    examples=[[ex, "ALBERT"] for ex in examples],
+    title="💸 Fraud Detection Assistant (ALBERT, FinBERT, CatBoost)",
+    description="Analyze transaction text for fraud using your choice of model."
+)
+
+# Launch
+if __name__ == "__main__":
+    gui.launch()