import streamlit as st
import pandas as pd
import numpy as np
import pickle
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestClassifier
import matplotlib.pyplot as plt

# Load the dictionary containing fitted scalers from the pickle file
with open('scalers.pkl', 'rb') as f:
    scalers = pickle.load(f)

# Access the individual scaler objects from the dictionary
fitted_scaler_amount = scalers['amount_scaler']
fitted_scaler_time = scalers['time_scaler']

# Load pre-trained model
with open('rf_random_model.pkl', 'rb') as model_file:
    model = pickle.load(model_file)

def predict_fraud(user_input):
    user_input_amount = user_input['Amount']
    user_input_time = user_input['Time']
    user_input_features = user_input.drop(columns=['Amount', 'Time'])

    # Scale the amount and time columns
    user_input_amount_scaled = fitted_scaler_amount.transform(np.array(user_input_amount).reshape(-1, 1))
    user_input_time_scaled = fitted_scaler_time.transform(np.array(user_input_time).reshape(-1, 1))

    # Reshape user_input_features if necessary
    if len(user_input_features.shape) == 1:
        user_input_features = user_input_features.values.reshape(1, -1)

    # Combine scaled amount and time with other features
    user_input_scaled = np.concatenate((user_input_features, user_input_amount_scaled, user_input_time_scaled), axis=1)

    # Ensure user_input_scaled is a 2D array
    if len(user_input_scaled.shape) == 1:
        user_input_scaled = user_input_scaled.reshape(1, -1)

    # Make predictions
    prediction = model.predict(user_input_scaled)[0]
    probability = model.predict_proba(user_input_scaled)[0][1]
    return prediction, probability

def generate_charts(prediction, probability, amount):
    fig, axes = plt.subplots(1, 2, figsize=(12, 6))

    # Pie chart for prediction distribution
    prediction_labels = ['Non-Fraudulent', 'Fraudulent']
    prediction_values = [1 - prediction, prediction]
    axes[0].pie(prediction_values, labels=prediction_labels, autopct='%1.1f%%', startangle=90, colors=['skyblue', 'lightcoral'])
    axes[0].set_title('Prediction Distribution')

    # Bar chart for probability distribution
    axes[1].bar(['Probability'], [probability], color='lightgreen' if prediction == 0 else 'lightcoral')
    axes[1].set_title('Probability of Fraud')

    # Display amount
    st.write(f"Transaction Amount: ${amount}")

    # Display charts
    st.pyplot(fig)


def main():
    # Set page layout
    st.set_page_config(
        page_title="Fraud Detection Prediction App",
        layout="wide",
        initial_sidebar_state="collapsed"
    )

    # Sidebar - User input fields
    st.sidebar.title("Input Parameters")
    user_input = {}
    for i in range(1, 29):
        user_input[f'V{i}'] = st.sidebar.number_input(f'V{i}', value=0.0)

    min_amount = st.sidebar.number_input('Minimum Amount', value=0.0, step=0.01)
    max_amount = st.sidebar.number_input('Maximum Amount', value=10000.0, step=0.01)
    amount = st.sidebar.number_input('Amount', min_value=min_amount, max_value=max_amount, value=(min_amount + max_amount) / 2, step=0.01)

    min_time = st.sidebar.number_input('Minimum Time', value=0.0, min_value=0.0, max_value=172800.0, step=1.0)
    max_time = st.sidebar.number_input('Maximum Time', value=172800.0, min_value=0.0, max_value=172800.0, step=1.0)
    time = st.sidebar.number_input('Time', min_value=min_time, max_value=max_time, value=(min_time + max_time) / 2, step=1.0)

    # Descriptions and explanations
    st.markdown("# Fraud Detection Prediction App")
    st.markdown("This app predicts the likelihood of a transaction being fraudulent based on input features.")
    st.markdown("The input features include V1 to V28, representing anonymized numerical features, and the transaction amount and time.")

    # Make prediction on button click
    if st.sidebar.button('Predict'):
        user_input_df = pd.DataFrame([user_input])  # Convert list of dictionaries to DataFrame
        user_input_df['Amount'] = amount
        user_input_df['Time'] = time
        prediction, probability = predict_fraud(user_input_df)

        # Display prediction result
        st.markdown("# Prediction Result")
        st.write(f"Prediction: {'Fraudulent Transaction' if prediction == 1 else 'Non-Fraudulent Transaction'}")
        st.write(f"Probability: {probability:.2%}")

        # Generate, display, and save charts
        generate_charts(prediction, probability, amount)


if __name__ == "__main__":
    main()