app.py

import gradio as gr
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import os
import uuid
import joblib

# Load the model
model = joblib.load("anomaly_detector_rf_model.pkl")

# Define the features expected by the model
expected_features = ["amount"]  # Update this list as per your trained model

def detect_anomalies(df):
    df = df.copy()
    df['is_anomalous'] = model.predict(df[expected_features])
    anomalies = df[df['is_anomalous'] == 1]

    # Save anomalies to temporary CSV file
    csv_filename = f"/tmp/anomalies_{uuid.uuid4().hex}.csv"
    anomalies.to_csv(csv_filename, index=False)

    return (
        anomalies[["transaction_id", "merchant", "location", "amount", "is_anomalous"]],
        csv_filename
    )

def generate_summary(df):
    total_transactions = len(df)
    total_anomalies = df['is_anomalous'].sum()
    percent_anomalies = round((total_anomalies / total_transactions) * 100, 2)
    return f"Total Transactions: {total_transactions}\nTotal Anomalies: {total_anomalies}\nAnomaly Rate: {percent_anomalies}%"

def generate_charts(df):
    fig, ax = plt.subplots(1, 2, figsize=(12, 5))

    # Distribution of Amounts
    sns.histplot(df['amount'], bins=30, ax=ax[0], kde=True)
    ax[0].set_title('Transaction Amount Distribution')

    # Anomalies by Merchant
    anomaly_counts = df[df['is_anomalous'] == 1]['merchant'].value_counts().nlargest(10)
    sns.barplot(x=anomaly_counts.values, y=anomaly_counts.index, ax=ax[1])
    ax[1].set_title('Top 10 Merchants with Anomalies')

    plt.tight_layout()
    chart_path = f"/tmp/chart_{uuid.uuid4().hex}.png"
    plt.savefig(chart_path)
    plt.close()
    return chart_path

def app_interface(file):
    df = pd.read_csv(file.name)
    anomalies, csv_path = detect_anomalies(df)
    summary = generate_summary(df)
    chart_path = generate_charts(df)
    return anomalies, summary, chart_path, csv_path

with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("""
    # 🔍 Elder Financial Abuse Detection Tool
    Upload a transaction dataset to identify potential financial abuse patterns in elderly individuals.
    """)

    with gr.Row():
        file_input = gr.File(label="📂 Upload Transaction CSV")
        analyze_btn = gr.Button("🚀 Analyze")

    with gr.Row():
        anomalies_output = gr.Dataframe(label="⚠️ Detected Anomalies", wrap=True)
        summary_output = gr.Textbox(label="📊 Summary")

    chart_output = gr.Image(label="📈 Analysis Charts")
    csv_download = gr.File(label="📁 Download Anomalies CSV")

    analyze_btn.click(
        fn=app_interface,
        inputs=[file_input],
        outputs=[anomalies_output, summary_output, chart_output, csv_download]
    )

demo.launch()