app.py

import pandas as pd
import numpy as np
from sklearn.linear_model import LinearRegression
from sklearn.ensemble import RandomForestRegressor
import gradio as gr
import yfinance as yf
import finnhub
import logging

# Setup logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')

# Initialize Finnhub client (for financial data)
finnhub_client = finnhub.Client(api_key="cr5lji9r01qgfrnl5q8gcr5lji9r01qgfrnl5q90")

# Define Shariah compliance rules
def is_shariah_compliant(row):
    try:
        # Check debt/asset ratio
        if row['debt_to_asset_ratio'] > 0.33:
            return False
        # Check interest income
        if row['interest_income'] / row['total_revenue'] > 0.05:
            return False
        # Check prohibited business activities
        if row['alcohol_revenue'] > 0 or row['gambling_revenue'] > 0 or row['tobacco_revenue'] > 0:
            return False
        return True
    except Exception as e:
        logging.error(f"Error in Shariah compliance check: {e}")
        return False

def get_stock_data(market):
    tickers = []
    if market == 'US':
        tickers = yf.Tickers("^GSPC").symbols
    elif market == 'HK':
        tickers = yf.Tickers("^HSI").symbols
    elif market == 'IN':
        tickers = yf.Tickers("^BSESN").symbols
    else:
        logging.warning(f"Market {market} is not supported")
        return pd.DataFrame()

    # Fetch financial data from Finnhub
    data = []
    for ticker in tickers:
        try:
            company_profile = finnhub_client.company_profile2(symbol=ticker)
            financials = finnhub_client.company_basic_financials(symbol=ticker, metric='all')
            
            row = {
                'stock_symbol': ticker,
                'market': market,
                'sector': company_profile.get('finnhubIndustry', 'Unknown'),
                'debt_to_asset_ratio': financials['metric'].get('debtToAssets', np.nan),
                'interest_income': financials['metric'].get('interestIncome', np.nan),
                'total_revenue': financials['metric'].get('totalRevenue', np.nan),
                'alcohol_revenue': 0,  # Placeholder, needs additional data source
                'gambling_revenue': 0,  # Placeholder, needs additional data source
                'tobacco_revenue': 0  # Placeholder, needs additional data source
            }
            data.append(row)
        except Exception as e:
            logging.error(f"Failed to fetch data for {ticker}: {e}")
            continue
    
    return pd.DataFrame(data)

def predict_stock_performance(market, sector, investment_horizon, model_type):
    # Fetch stock data for the selected market
    df = get_stock_data(market)
    if df.empty:
        return "No data available for the selected market."
    
    # Apply Shariah compliance filtering
    df['is_shariah_compliant'] = df.apply(is_shariah_compliant, axis=1)
    shariah_compliant_df = df[df['is_shariah_compliant']]
    
    if shariah_compliant_df.empty:
        return "No Shariah-compliant stocks found."

    # Check for required columns before training the model
    required_columns = ['price_to_earnings', 'dividend_yield', 'revenue_growth']
    for col in required_columns:
        if col not in shariah_compliant_df.columns:
            logging.error(f"Missing column: {col}")
            return f"Error: Missing required financial data ({col})."
    
    # Financial analysis and forecasting
    X = shariah_compliant_df[required_columns]
    y = shariah_compliant_df.get('future_price', pd.Series())

    if y.empty:
        return "Error: Future price data not available."

    # Train the selected model
    try:
        if model_type == 'linear_regression':
            model = LinearRegression()
        elif model_type == 'random_forest':
            model = RandomForestRegressor()
        else:
            return "Invalid model type selected."

        model.fit(X, y)

        # Predict future stock performance
        shariah_compliant_df['predicted_price'] = model.predict(X)
    except Exception as e:
        logging.error(f"Model training failed: {e}")
        return "Error in model training or prediction."

    # Rank and recommend the top Shariah-compliant stocks
    top_stocks = shariah_compliant_df[shariah_compliant_df['sector'] == sector].sort_values('predicted_price', ascending=False).head(5)
    
    if top_stocks.empty:
        return "No top stocks found in the selected sector."

    # Prepare the output for the Gradio interface
    output = '\n'.join(f"{row['stock_symbol']}: {row['predicted_price']}" for _, row in top_stocks.iterrows())
    return output

def update_sectors(market):
    df = get_stock_data(market)
    if df.empty:
        return []

    return df['sector'].unique().tolist()

# Create the Gradio interface
with gr.Blocks() as app:
    with gr.Row():
        market = gr.Dropdown(['US', 'HK', 'IN'], label='Market', interactive=True)
        sector = gr.Dropdown([], label='Sector', interactive=True)
    
    investment_horizon = gr.Slider(1, 10, step=1, label='Investment Horizon (years)')
    model_type = gr.Radio(['linear_regression', 'random_forest'], label='Model Type')

    output = gr.Textbox(label='Recommended Shariah-Compliant Stocks')

    market.change(fn=update_sectors, inputs=[market], outputs=[sector])
    gr.Button('Predict').click(fn=predict_stock_performance, inputs=[market, sector, investment_horizon, model_type], outputs=[output])

app.launch()