app.py

import streamlit as st
import pandas as pd
import numpy as np
import pickle
import os
from data_processing import load_sample_data, preprocess_inputs
from model_utils import load_model, predict_price

# Set page config
st.set_page_config(
    page_title="House Price Predictor",
    page_icon="🏠",
    layout="wide"
)

# Load custom CSS
def load_css(css_file):
    try:
        # Essayer avec UTF-8 explicitement
        with open(css_file, 'r', encoding='utf-8') as f:
            css = f.read()
        return css
    except UnicodeDecodeError:
        # Fallback sur une autre encodage si UTF-8 échoue
        try:
            with open(css_file, 'r', encoding='latin-1') as f:
                css = f.read()
            return css
        except Exception as e:
            return None

try:
    css = load_css('style.css')
    if css:
        st.markdown(f'<style>{css}</style>', unsafe_allow_html=True)
    else:
        # Si le fichier CSS ne peut pas être chargé, ajouter un style minimal inline
        st.markdown("""
        <style>
        h1 {
            font-size: 2rem;
            color: #7f8c8d;
            text-align: center;
            margin-bottom: 1rem;
        }
        .sub-header {
            font-size: 1.2rem;
            color: #7f8c8d;
            text-align: center;
            margin-bottom: 2rem;
        }
        .feature-section {
            background-color: white;
            padding: 20px;
            border-radius: 10px;
            box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
            margin-bottom: 20px;
        }
        .prediction-result {
            font-size: 1.5rem;
            font-weight: bold;
            text-align: center;
            padding: 20px;
            border-radius: 10px;
            margin: 20px 0;
            background: linear-gradient(to right, #3498db, #2c3e50);
            color: white;
        }
        </style>
        """, unsafe_allow_html=True)
except Exception as e:
    st.write(f"Erreur lors du chargement du CSS (mode dégradé activé): {e}")

# Custom styling for specific components
st.markdown("""
    <style>
    h1 {
        font-size: 2.5rem;
        color: #D4DCFF;
        text-align: center;
        margin-bottom: 1rem;
    }
    .sub-header {
        font-size: 1.2rem;
        color: #D4DCFF;
        text-align: center;
        margin-bottom: 2rem;
        font-style: italic;
    }
    .feature-section {
        color: #D4DCFF:
        padding: 20px;
        border-radius: 10px;
        box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
        margin-bottom: 20px;
    }
    .prediction-result {
        font-size: 1.5rem;
        font-weight: bold;
        text-align: center;
        padding: 20px;
        border-radius: 10px;
        margin: 20px 0;
        background: linear-gradient(to right, #3498db, #2c3e50);
        color: white;
    }
    .confidence-interval {
        font-size: 1.1rem;
        text-align: center;
        color: #7f8c8d;
        margin-top: -10px;
        margin-bottom: 20px;
    }
    .footer {
        text-align: center;
        color: #95a5a6;
        padding: 20px;
        font-size: 0.8rem;
    }
    </style>
    """, unsafe_allow_html=True)

# Title and description with custom styling
st.markdown('<h1>🏠 Prédiction de Prix Immobilier</h1>', unsafe_allow_html=True)
st.markdown('<p class="sub-header">Cette application vous permet de prédire le prix des propriétés en utilisant un modèle XGBoost pré-entraîné.</p>', unsafe_allow_html=True)

# Load sample data for statistics
sample_data = load_sample_data()

# Sidebar for model upload and data statistics
with st.sidebar:
    st.markdown("""
    <style>
    .sidebar-header {
        background-color: #171933;
        color: #D4DCFF;
        border-radius: 5px;
        text-align: center;
        font-weight: bold;
    }
    .sidebar-box {
        border-radius: 8px;
        margin-bottom: 20px;
    }
    .sidebar-subheader {
        background-color: #171933;
        color: #2c3e50;
        margin-top: 20px;
        font-weight: 600;
        text-align: center;
        border-radius: 5px;
        margin-bottom: 20px;
    }
    .stat-item {
        background-color: #171933;
        padding: 8px;
        border-radius: 5px;
        margin-bottom: 5px;
        border-left: 3px solid #02b35a;
    }
    </style>
    """, unsafe_allow_html=True)
    
    # Model upload section
    st.markdown('<h2 class="sidebar-header">Télécharger le Modèle pour faire des prédictions</h2>', unsafe_allow_html=True)
    
    st.markdown('<div class="sidebar-box">', unsafe_allow_html=True)
    uploaded_model = st.file_uploader("", type=["pkl", "joblib", "sav"])
    
    if uploaded_model is not None:
        st.success("✅ Modèle téléchargé avec succès!")
    
    # Confidence interval settings
    st.markdown('<h2 class="sidebar-header">⚙️ Paramètres</h2>', unsafe_allow_html=True)
    st.markdown('<div style="color: #D4DCFF;" class="sidebar-box">', unsafe_allow_html=True)
    
    # Add confidence level settings
    confidence_level = st.slider(
        "Niveau de confiance (%)", 
        min_value=50, 
        max_value=99, 
        value=95, 
        step=5, 
        help="Niveau de confiance pour l'intervalle de prédiction"
    )
    
    # Add error margin slider
    error_margin_percent = st.slider(
        "Marge d'erreur estimée (%)", 
        min_value=5, 
        max_value=30, 
        value=15, 
        step=5,
        help="Pourcentage d'erreur estimée pour le modèle"
    )
    
    st.markdown('</div>', unsafe_allow_html=True)
    
    # Dataset Statistics
    
    if sample_data is not None:
        # Preview
        st.markdown('<div class="sidebar-box">', unsafe_allow_html=True)
        st.markdown('<h3 style="margin-bottom: 15px;" class="sidebar-subheader">📊 Aperçu du Dataset</h3>', unsafe_allow_html=True)
        st.dataframe(sample_data.head(3), use_container_width=True)
        st.markdown('</div>', unsafe_allow_html=True)
        
        # Features stats
        st.markdown('<div class="sidebar-box">', unsafe_allow_html=True)
        st.markdown('<h3 class="sidebar-subheader">Caractéristiques Numériques</h3>', unsafe_allow_html=True)
        
        # Price stats (highlighted)
        if 'price' in sample_data.columns:
            price_stats = sample_data['price'].describe()
            st.markdown("""
            <div style="background-color: #171933; padding: 10px; border-radius: 5px; margin-bottom: 15px; margin-top: 15px; border-left: 3px solid #3498db;">
                <h4 style="margin: 0; color: #D4DCFF;">Prix (Target Variable)</h4>
                <p style="margin: 5px 0;">Min: ${:,.2f} | Max: ${:,.2f}</p>
                <p style="margin: 5px 0;">Mean: ${:,.2f} | Median: ${:,.2f}</p>
            </div>
            """.format(
                price_stats['min'], 
                price_stats['max'], 
                price_stats['mean'], 
                price_stats['50%']
            ), unsafe_allow_html=True)
        
        # Other features
        num_cols = sample_data.select_dtypes(include=['int64', 'float64']).columns
        for col in num_cols:
            if col != 'id' and col != 'price':  # Exclude ID and target variable
                stats = sample_data[col].describe()
                st.markdown(f"""
                <div class="stat-item">
                    <strong>{col}</strong><br/>
                    Min: {stats['min']:.2f} | Max: {stats['max']:.2f}<br/>
                    Mean: {stats['mean']:.2f}
                </div>
                """, unsafe_allow_html=True)
        
        st.markdown('</div>', unsafe_allow_html=True)

# Main content
st.markdown('<h2 style="color: #D4DCFF; padding-bottom: 5px;">Détails de la Propriété</h2>', unsafe_allow_html=True)

# Initialize model
model = None
if uploaded_model is not None:
    try:
        model = load_model(uploaded_model)
        # Success message is now shown in the sidebar
    except Exception as e:
        st.error(f"Erreur lors du chargement du modèle : {e}")
        st.info("Veuillez vérifier que le modèle est un fichier XGBoost valide au format .pkl")

# Define all available features
all_features = {
    'bedrooms': "Bedrooms",
    'bathrooms': "Bathrooms", 
    'sqft_living': "Living Area (sqft)",
    'sqft_lot': "Lot Size (sqft)",
    'floors': "Floors",
    'waterfront': "Waterfront",
    'view': "View",
    'condition': "Condition",
    'grade': "Grade",
    'yr_built': "Year Built",
    'yr_renovated': "Year Renovated",
    'zipcode': "Zipcode",
    'lat': "Latitude",
    'long': "Longitude",
    'sqft_above': "Square Feet Above Ground",
    'sqft_basement': "Square Feet Basement"
}

# Multi-select to choose fields to omit - styling with custom container
st.markdown('<div class="feature-section">', unsafe_allow_html=True)
st.subheader("Sélectionnez les champs à omettre pour la prédiction")
omitted_fields = st.multiselect(
    "",
    options=list(all_features.values())
)
st.markdown('</div>', unsafe_allow_html=True)

# Create lookup from display name to field name
display_to_field = {v: k for k, v in all_features.items()}

# Get the field names that are omitted
omitted_field_names = [display_to_field[field] for field in omitted_fields]

# Create form for user inputs
with st.form("property_form"):
    # Apply CSS to the form
    st.markdown("""
    <style>
    .property-form {
        background-color: #22244D;
    }            
    .feature-label {
        font-weight: 600;
        color: #D4DCFF;
        margin-bottom: 5px;
    }
    .submit-btn {
        background-color: #3498db;
        color: white;
        font-weight: bold;
        padding: 10px 20px;
        border-radius: 5px;
        border: none;
        cursor: pointer;
        transition: all 0.3s ease;
        display: block;
        width: 100%;
        text-align: center;
        margin-top: 20px;
    }
    .submit-btn:hover {
        background-color: #2980b9;
        box-shadow: 0 2px 5px rgba(0, 0, 0, 0.2);
    }
    </style>
    """, unsafe_allow_html=True)

    # Wrap the form in a styled container
    st.markdown('<div class="property-form">', unsafe_allow_html=True)
    
    # Create columns for a cleaner form layout
    col1, col2, col3 = st.columns(3)
    
    # Initialize input variables with None
    bedrooms = bathrooms = sqft_living = sqft_lot = floors = None
    waterfront = view = condition = grade = yr_built = None
    yr_renovated = zipcode = lat = long = sqft_above = sqft_basement = None
    
    # Column 1 features
    with col1:
        if "Bedrooms" not in omitted_fields:
            st.markdown('<div class="feature-label">Chambres</div>', unsafe_allow_html=True)
            bedrooms = st.number_input("Chambres", min_value=0, max_value=10, value=3, step=1, label_visibility="collapsed")
        
        if "Bathrooms" not in omitted_fields:
            st.markdown('<div class="feature-label">Salles de bain</div>', unsafe_allow_html=True)
            bathrooms = st.number_input("Salles de bain", min_value=0.0, max_value=10.0, value=2.0, step=0.25, label_visibility="collapsed")
        
        if "Living Area (sqft)" not in omitted_fields:
            st.markdown('<div class="feature-label">Surface habitable (pieds carrés)</div>', unsafe_allow_html=True)
            sqft_living = st.number_input("Surface habitable", min_value=200, max_value=10000, value=1500, step=100, label_visibility="collapsed")
        
        if "Lot Size (sqft)" not in omitted_fields:
            st.markdown('<div class="feature-label">Taille du lot (pieds carrés)</div>', unsafe_allow_html=True)
            sqft_lot = st.number_input("Taille du lot", min_value=500, max_value=200000, value=5000, step=500, label_visibility="collapsed")
        
        if "Floors" not in omitted_fields:
            st.markdown('<div class="feature-label">Étages</div>', unsafe_allow_html=True)
            floors = st.number_input("Étages", min_value=1.0, max_value=4.0, value=1.0, step=0.5, label_visibility="collapsed")
    
    # Column 2 features
    with col2:
        if "Waterfront" not in omitted_fields:
            st.markdown('<div class="feature-label">Vue sur l\'eau</div>', unsafe_allow_html=True)
            waterfront = st.selectbox("Vue sur l'eau", options=[0, 1], 
                                    format_func=lambda x: "Oui" if x == 1 else "Non", 
                                    label_visibility="collapsed")
        
        if "View" not in omitted_fields:
            st.markdown('<div class="feature-label">Vue</div>', unsafe_allow_html=True)
            view = st.selectbox("Vue", options=[0, 1, 2, 3, 4], 
                              format_func=lambda x: {0: "Aucune", 1: "Passable", 2: "Moyenne", 3: "Bonne", 4: "Excellente"}[x],
                              label_visibility="collapsed")
        
        if "Condition" not in omitted_fields:
            st.markdown('<div class="feature-label">État</div>', unsafe_allow_html=True)
            condition = st.selectbox("État", options=[1, 2, 3, 4, 5], 
                                   format_func=lambda x: {1: "Mauvais", 2: "Passable", 3: "Moyen", 4: "Bon", 5: "Excellent"}[x],
                                   label_visibility="collapsed")
        
        if "Grade" not in omitted_fields:
            st.markdown('<div class="feature-label">Qualité de construction</div>', unsafe_allow_html=True)
            grade = st.selectbox("Qualité", options=list(range(1, 14)), 
                               format_func=lambda x: {
                                   1: "Très mauvaise", 2: "Mauvaise", 3: "Mauvaise", 4: "Moyenne inférieure", 
                                   5: "Moyenne", 6: "Moyenne", 7: "Bonne", 8: "Bonne", 
                                   9: "Meilleure", 10: "Meilleure", 11: "Excellente", 
                                   12: "Excellente", 13: "Luxe"
                               }[x],
                               label_visibility="collapsed")
        
        if "Year Built" not in omitted_fields:
            st.markdown('<div class="feature-label">Année de construction</div>', unsafe_allow_html=True)
            yr_built = st.number_input("Année de construction", min_value=1900, max_value=2023, value=1980, step=1, label_visibility="collapsed")
    
    # Column 3 features
    with col3:
        if "Year Renovated" not in omitted_fields:
            st.markdown('<div class="feature-label">Année de rénovation (0 si aucune)</div>', unsafe_allow_html=True)
            yr_renovated = st.number_input("Année de rénovation", min_value=0, max_value=2023, value=0, step=1, label_visibility="collapsed")
        
        if "Zipcode" not in omitted_fields:
            st.markdown('<div class="feature-label">Code postal</div>', unsafe_allow_html=True)
            zipcode = st.number_input("Code postal", min_value=98000, max_value=99000, value=98000, step=1, label_visibility="collapsed")
        
        if "Latitude" not in omitted_fields:
            st.markdown('<div class="feature-label">Latitude</div>', unsafe_allow_html=True)
            lat = st.number_input("Latitude", min_value=47.0, max_value=48.0, value=47.5, step=0.01, format="%.4f", label_visibility="collapsed")
        
        if "Longitude" not in omitted_fields:
            st.markdown('<div class="feature-label">Longitude</div>', unsafe_allow_html=True)
            long = st.number_input("Longitude", min_value=-123.0, max_value=-121.0, value=-122.0, step=0.01, format="%.4f", label_visibility="collapsed")
        
        if "Square Feet Above Ground" not in omitted_fields:
            st.markdown('<div class="feature-label">Surface au-dessus du sol (pieds carrés)</div>', unsafe_allow_html=True)
            sqft_above = st.number_input("Surface au-dessus du sol", min_value=200, max_value=10000, value=1000, step=100, label_visibility="collapsed")
        
        if "Square Feet Basement" not in omitted_fields:
            st.markdown('<div class="feature-label">Surface du sous-sol (pieds carrés)</div>', unsafe_allow_html=True)
            sqft_basement = st.number_input("Surface du sous-sol", min_value=0, max_value=5000, value=0, step=100, label_visibility="collapsed")
    
    # Close the container div
    st.markdown('</div>', unsafe_allow_html=True)
    
    # Submit button with custom styling
    col1, col2, col3 = st.columns([1, 2, 1])
    with col2:
        submitted = st.form_submit_button("Prédire le Prix", use_container_width=True)
        # Add additional styling to the button
        st.markdown("""
        <style>
        div.stButton > button {
            background-color: #3498db;
            color: white;
            font-weight: bold;
            border-radius: 5px;
            border: none;
            padding: 0.5em 1em;
            font-size: 1.2em;
        }
        div.stButton > button:hover {
            background-color: #2980b9;
            box-shadow: 0 2px 5px rgba(0, 0, 0, 0.2);
        }
        </style>
        """, unsafe_allow_html=True)

# Process the form submission
if submitted:
    if model is None:
        st.error("Veuillez télécharger un modèle d'abord.")
    else:
        try:
            # Get the confidence level and error margin from the sidebar
            conf_level = confidence_level / 100.0
            error_margin = error_margin_percent / 100.0
            
            # Prepare input data
            input_data = {
                'bedrooms': bedrooms,
                'bathrooms': bathrooms,
                'sqft_living': sqft_living,
                'sqft_lot': sqft_lot,
                'floors': floors,
                'waterfront': waterfront,
                'view': view,
                'condition': condition,
                'grade': grade,
                'sqft_above': sqft_above,
                'sqft_basement': sqft_basement,
                'yr_built': yr_built,
                'yr_renovated': yr_renovated,
                'zipcode': zipcode,
                'lat': lat,
                'long': long,
                'sqft_living15': sqft_living,  # Using the same value as sqft_living as an approximation
                'sqft_lot15': sqft_lot  # Using the same value as sqft_lot as an approximation
            }
            
            # Track omitted features
            omitted_features = {k: None for k, v in input_data.items() if v is None}
            
            # Preprocess inputs
            processed_input = preprocess_inputs(input_data)
            
            # Make prediction with confidence interval
            predicted_price, interval_lower, interval_upper = predict_price(
                model, 
                processed_input, 
                with_confidence_interval=True, 
                confidence_level=conf_level, 
                error_margin=error_margin
            )
            
            # Display prediction with custom styling
            st.markdown(f'<div class="prediction-result">Prix Prédit: ${predicted_price:,.2f}</div>', unsafe_allow_html=True)
            
            # Display confidence interval
            st.markdown(f"""
            <div class="confidence-interval">
                Intervalle de confiance ({confidence_level}%): ${interval_lower:,.2f} - ${interval_upper:,.2f}
            </div>
            """, unsafe_allow_html=True)
            
            # Show input summary with styled container
            with st.expander("📋 Résumé des détails de la propriété"):
                st.markdown('<div class="feature-section">', unsafe_allow_html=True)
                
                # Show which features were used as provided by user
                st.subheader("🔹 Caractéristiques fournies par l'utilisateur")
                provided_features = {k: v for k, v in input_data.items() if v is not None}
                if provided_features:
                    st.json(provided_features)
                else:
                    st.info("Aucune caractéristique n'a été fournie par l'utilisateur.")
                
                # Show which features were omitted and filled with default values
                if omitted_features:
                    st.subheader("🔸 Caractéristiques omises (remplies avec des valeurs par défaut)")
                    st.write("Les caractéristiques suivantes ont été omises et remplies avec des valeurs par défaut du jeu de données:")
                    
                    # Create columns for better display
                    omitted_cols = st.columns(3)
                    for i, feature in enumerate(omitted_features):
                        col_idx = i % 3
                        with omitted_cols[col_idx]:
                            st.markdown(f"**{feature}**: {processed_input[feature].values[0]:.2f}")
                
                # Show the final processed input used for prediction
                st.subheader("📊 Données d'entrée finales utilisées")
                st.dataframe(processed_input, use_container_width=True)
                
                # Show confidence interval details
                st.subheader("📈 Détails de l'intervalle de confiance")
                st.write(f"Niveau de confiance: {confidence_level}%")
                st.write(f"Marge d'erreur estimée: {error_margin_percent}%")
                st.write(f"Intervalle: ${interval_lower:,.2f} - ${interval_upper:,.2f}")
                
                st.markdown('</div>', unsafe_allow_html=True)
                
        except Exception as e:
            st.error(f"Une erreur s'est produite lors de la prédiction: {e}")
            st.error("Détails: " + str(e))  # More detailed error message


# Add a footer
st.markdown('<div class="footer">Développé avec BONDA DENICLO Emilio | © 2025</div>', unsafe_allow_html=True)