app.py

import time
import streamlit as st
import pandas as pd
import matplotlib.pyplot as plt
import plotly.express as px

# Set page configuration
st.set_page_config(
    page_title="Enhanced Fraud Detection Demo",
    page_icon="🔍",
    layout="wide",
    initial_sidebar_state="expanded",
)

# Title of the app
st.title("🔍 Enhanced Fraud Detection Simulation Demo")

# Security and Compliance Information
st.markdown("""
---
**Security & Compliance:**
- **Data Protection:** All data is encrypted and securely stored.
- **Access Control:** Role-based access control implemented for all users.
---
""")

# Sidebar for user inputs
st.sidebar.header("Simulate Submission Data")
applicant_id = st.sidebar.text_input("Applicant ID", "A12345")
name = st.sidebar.text_input("Name", "John Doe")
age = st.sidebar.number_input("Age", 18, 100, 34)
income = st.sidebar.number_input("Income ($)", 0, 10000, 2500)
benefits = st.sidebar.multiselect(
    "Benefits Applied For",
    ["Food Assistance", "Child Support", "Medical Coverage", "Temporary Assistance"],
    ["Food Assistance", "Child Support"]
)
address = st.sidebar.text_input("Address", "123 Elm Street")
ebt_card = st.sidebar.text_input("EBT Card Number", "EBT123456789")
app_date = st.sidebar.date_input("Application Date", value=pd.to_datetime("2024-10-25"))

# Scenario selection
scenario = st.sidebar.selectbox(
    "Select Submission Scenario",
    ["Custom Submission", "Normal Submission", "Potential Fraud"]
)

# Adjust submission data based on scenario
if scenario == "Normal Submission":
    income = 5000
    benefits = ["Food Assistance"]
    address_change = False
elif scenario == "Potential Fraud":
    income = 2500
    benefits = ["Food Assistance", "Child Support"]
    address_change = True
else:
    # Custom submission
    address_change = True if "Elm Street" in address else False

# Help Section
with st.expander("ℹ️ Help & Documentation"):
    st.markdown("""
    **How to Use This Demo:**
    1. **Simulate Submission:** Enter applicant details in the sidebar or select a predefined scenario.
    2. **Start Simulation:** Click the "Start Fraud Detection Simulation" button to begin.
    3. **View Results:** The app will display the submission data, anomaly detection results, a detailed report, and visualizations.
    4. **Rerun Simulation:** Use the "Rerun Simulation" button to test different scenarios.

    **Understanding the Outputs:**
    - **Anomaly Score:** A higher score indicates a higher likelihood of fraud.
    - **Fraud Analysis Report:** Provides detailed reasons and recommendations based on detected anomalies.
    - **Visualizations:** Graphical representations of anomaly scores and distributions.

    **Security Features:**
    - All data is handled securely and complies with relevant standards.
    """)

# Button to start the simulation
if st.button("Start Fraud Detection Simulation"):
    # Step 1: Submission Intake
    with st.status("Receiving submission..."):
        st.write("📥 **Submission Received:**")
        # Simulated submission data
        submission_data = {
            "Applicant ID": applicant_id,
            "Name": name,
            "Age": age,
            "Income": income,
            "Benefits Applied For": benefits,
            "Address": address,
            "EBT Card Number": ebt_card,
            "Application Date": str(app_date)
        }
        st.json(submission_data)
        progress = st.progress(20)
        time.sleep(1)
        progress.progress(40)
        time.sleep(1)

    # Simulate fetching additional data from EBT vendor (FIS)
    with st.status("Fetching data from EBT vendor (FIS)..."):
        st.write("🔗 **Connecting to EBT Vendor (FIS):**")
        # Simulated external data
        ebt_vendor_data = {
            "EBT Card Number": "EBT123456789",
            "Balance": 1500,
            "Last Transaction": "2024-10-24",
            "Transaction History": [
                {"Date": "2024-10-20", "Amount": 100, "Merchant": "SuperMart"},
                {"Date": "2024-10-22", "Amount": 200, "Merchant": "QuickStop"},
            ]
        }
        st.json(ebt_vendor_data)
        progress.progress(60)
        time.sleep(1)
        progress.progress(80)
        time.sleep(1)

    # Step 2: Anomaly Detection with ML
    with st.status("Analyzing submission for anomalies..."):
        st.write("🤖 **Running ML Anomaly Detection:**")
        
        # Simulated pattern data
        existing_submissions = pd.DataFrame([
            {"Applicant ID": "A12346", "Name": "Jane Smith", "Income": 5200, "Benefits": ["Food Assistance"], "Address": "456 Oak Street", "EBT Card Number": "EBT987654321"},
            {"Applicant ID": "A12347", "Name": "Bob Johnson", "Income": 3000, "Benefits": ["Child Support"], "Address": "789 Pine Street", "EBT Card Number": "EBT192837465"},
            # Add more submissions as needed
        ])

        terminal_usage = pd.DataFrame([
            {"Terminal ID": "T001", "Location": "SuperMart Downtown", "Anomaly Score": 0.1},
            {"Terminal ID": "T002", "Location": "QuickStop Uptown", "Anomaly Score": 0.3},
            {"Terminal ID": "T003", "Location": "MegaStore Central", "Anomaly Score": 0.7},
            # Add more terminals as needed
        ])

        call_center_patterns = pd.DataFrame([
            {"Call Center ID": "C001", "Activity": "High volume of calls", "Anomaly Score": 0.4},
            {"Call Center ID": "C002", "Activity": "Repeated failed verifications", "Anomaly Score": 0.6},
            {"Call Center ID": "C003", "Activity": "Unusual call times", "Anomaly Score": 0.2},
            # Add more patterns as needed
        ])

        # Simple anomaly detection logic across patterns
        def detect_patterns(income, benefits, address_change, ebt_balance, terminal_id, call_center_id):
            anomaly_score = 0
            reasons = []

            # Income anomaly
            if income < 3000:
                anomaly_score += 0.2
                reasons.append("Income significantly lower than average for applicants.")

            # Multiple benefits
            if len(benefits) > 1:
                anomaly_score += 0.2
                reasons.append("Multiple benefits applied for simultaneously.")

            # Address change
            if address_change:
                anomaly_score += 0.1
                reasons.append("Recent change in address detected.")

            # EBT balance anomaly
            if ebt_balance < 500:
                anomaly_score += 0.2
                reasons.append("EBT balance unusually low.")

            # Terminal anomaly
            terminal = terminal_usage[terminal_usage["Terminal ID"] == terminal_id]
            if not terminal.empty and terminal["Anomaly Score"].values[0] > 0.5:
                anomaly_score += 0.2
                reasons.append(f"Terminal {terminal_id} at {terminal['Location'].values[0]} shows high anomaly score.")

            # Call center pattern anomaly
            call_center = call_center_patterns[call_center_patterns["Call Center ID"] == call_center_id]
            if not call_center.empty and call_center["Anomaly Score"].values[0] > 0.5:
                anomaly_score += 0.1
                reasons.append(f"Call Center {call_center_id} shows suspicious activity: {call_center['Activity'].values[0]}.")

            # Insider threat simulation (e.g., unusual internal access)
            insider_threat = False  # Hardcoded for simplicity
            if insider_threat:
                anomaly_score += 0.1
                reasons.append("Insider threat detected: Unusual internal access patterns.")

            return round(anomaly_score, 2), reasons

        # Simulate terminal and call center IDs
        simulated_terminal_id = "T003"  # High anomaly score terminal
        simulated_call_center_id = "C002"  # High anomaly score call center

        # EBT balance from vendor data
        ebt_balance = ebt_vendor_data["Balance"]

        # Calculate anomaly score
        anomaly_score, reasons = detect_patterns(
            income=income,
            benefits=benefits,
            address_change=address_change,
            ebt_balance=ebt_balance,
            terminal_id=simulated_terminal_id,
            call_center_id=simulated_call_center_id
        )

        ml_response = {
            "Anomaly Detected": anomaly_score > 0.5,
            "Anomaly Score": anomaly_score,
            "Reasons": reasons
        }

        # Simulate processing steps
        st.write("🔍 **Evaluating applicant data and detecting patterns...")
        time.sleep(1)
        st.write("📊 **Calculating comprehensive anomaly score...")
        time.sleep(1)
        st.write("✅ **Anomaly Detection Complete.")
        st.json(ml_response)
        progress.progress(100)
        time.sleep(0.5)

    # Step 3: Report Generation with AI
    with st.status("Generating fraud analysis report..."):
        st.write("📝 **Generating Report:**")
        
        # Generate AI report based on ML response
        def generate_report(applicant_id, name, anomaly_score, reasons):
            report = f"""
### Fraud Analysis Report

**Applicant ID:** {applicant_id}  
**Name:** {name}  
**Anomaly Score:** {anomaly_score}

**Reasons for Potential Fraud:**
"""
            for reason in reasons:
                report += f"- **{reason}**\n"

            report += """
**Recommendations:**
- Conduct a detailed investigation of the applicant's financial records.
- Verify the authenticity of the provided documents.
- Monitor EBT card transactions for any suspicious activities.
- Review terminal usage patterns and secure high-risk terminals.
- Investigate call center activities associated with the applicant.
"""
            return report

        ai_report = generate_report(applicant_id, name, ml_response["Anomaly Score"], ml_response["Reasons"])

        # Simulate report generation
        st.markdown(ai_report)
        time.sleep(1)

    # Step 4: Visualization
    with st.status("Creating visualization of detected anomalies..."):
        st.write("📈 **Generating Visualization:**")
        # Simulated data for visualization

        # Anomaly Scores by Category
        data = {
            "Category": ["Income", "Benefits Applied", "Address Changes", "EBT Transactions", "Terminal Usage", "Call Center Activity"],
            "Score": [0.2, 0.2, 0.1, 0.2, 0.2, 0.1]
        }
        df = pd.DataFrame(data)

        # Create a bar chart with Matplotlib
        fig, ax = plt.subplots(figsize=(10, 5))
        bars = ax.bar(df["Category"], df["Score"], color=['red', 'orange', 'yellow', 'red', 'orange', 'yellow'])
        ax.set_ylim(0, 1)
        ax.set_ylabel("Anomaly Score")
        ax.set_title("Anomaly Scores by Category")

        # Add score labels on top of bars
        for bar in bars:
            height = bar.get_height()
            ax.annotate(f'{height:.2f}',
                        xy=(bar.get_x() + bar.get_width() / 2, height),
                        xytext=(0, 3),  # 3 points vertical offset
                        textcoords="offset points",
                        ha='center', va='bottom')

        st.pyplot(fig)
        time.sleep(1)

        # Additional Visualization: Anomaly Score Distribution
        st.write("📊 **Anomaly Score Distribution:**")
        fig_pie = px.pie(
            values=[1 - anomaly_score, anomaly_score],
            names=["Normal", "Anomaly"],
            title="Anomaly Score Distribution",
            color_discrete_sequence=["#4CAF50", "#F44336"]
        )
        st.plotly_chart(fig_pie)
        time.sleep(1)

        # Terminal Anomaly Heatmap
        st.write("📍 **Terminal Anomaly Heatmap:**")
        # Simulated terminal locations and anomaly scores
        terminal_map_data = terminal_usage.copy()
        fig_map = px.scatter_geo(
            terminal_map_data,
            lat=[47.6062, 47.6097, 47.6205],  # Example latitudes
            lon=[-122.3321, -122.3416, -122.3493],  # Example longitudes
            text=terminal_map_data["Location"],
            size=terminal_map_data["Anomaly Score"] * 20,
            color=terminal_map_data["Anomaly Score"],
            color_continuous_scale="Reds",
            title="Terminal Anomaly Heatmap"
        )
        st.plotly_chart(fig_map)
        time.sleep(1)

    # Final Status
    st.success("✅ **Fraud Detection Simulation Completed Successfully!**")

    # Feedback Section
    with st.expander("💬 Provide Feedback"):
        feedback = st.text_area("We value your feedback. Please share your thoughts on this simulation.")
        if st.button("Submit Feedback"):
            st.success("Thank you for your feedback!")

    # Option to rerun
    st.markdown("---")
    st.button("Rerun Simulation")

else:
    st.info("Click the **Start Fraud Detection Simulation** button to begin.")