app.py

import h2o
import pandas as pd
import gradio as gr

# Initialize H2O server
h2o.init()

# Load the saved H2O model
saved_model_path = "XGBoost_1_AutoML_1_20240917_65817"  # Replace with your model path
model = h2o.load_model(saved_model_path)

# Define the prediction function
def predict_fraud(Amount, Use_Chip, Merchant_City, Merchant_State, MCC, Errors, FICO_Score, Card_on_Dark_Web, Num_Credit_Cards):
    # Create a pandas DataFrame from the input data
    data = {
        "Amount": [Amount],
        "Use_Chip": [Use_Chip],
        "Merchant_City": [Merchant_City],
        "Merchant_State": [Merchant_State],
        "MCC": [MCC],
        "Errors": [Errors],
        "FICO_Score": [FICO_Score],
        "Card_on_Dark_Web": [Card_on_Dark_Web],
        "Num_Credit_Cards": [Num_Credit_Cards]
    }
    df = pd.DataFrame(data)

    # Convert to H2OFrame
    hf = h2o.H2OFrame(df)

    # Get predictions from the model
    predictions = model.predict(hf)

    # Extract the prediction result
    return predictions.as_data_frame().iloc[0, 0]

# Create Gradio interface
interface = gr.Interface(
    fn=predict_fraud,
    inputs=[
        gr.inputs.Number(label="Amount"),
        gr.inputs.Dropdown(["Yes", "No"], label="Use Chip"),
        gr.inputs.Textbox(label="Merchant City"),
        gr.inputs.Textbox(label="Merchant State"),
        gr.inputs.Number(label="MCC"),
        gr.inputs.Textbox(label="Errors"),
        gr.inputs.Number(label="FICO Score"),
        gr.inputs.Dropdown(["Yes", "No"], label="Card on Dark Web"),
        gr.inputs.Number(label="Number of Credit Cards")
    ],
    outputs="text",
    title="Fraud Detection Model",
    description="Enter transaction details to predict if it's fraudulent."
)

# Launch the interface
interface.launch()