app.py

"""
House Price Predictor - HuggingFace Space
Standalone Gradio application for Seattle/King County house price prediction
"""
import gradio as gr
import pandas as pd
import plotly.express as px
import joblib
from pathlib import Path

# Paths (files in same directory as app.py)
MODEL_PATH = Path("house_price_model.joblib")
DATA_PATH = Path("kc_house_data.csv")

class HousePricePredictor:
    def __init__(self):
        self.model = joblib.load(MODEL_PATH)

    def predict(self, bedrooms, bathrooms, sqft, age):
        X = pd.DataFrame([[bedrooms, bathrooms, sqft, age]],
                        columns=['bedrooms', 'bathrooms', 'sqft', 'age'])
        price = self.model.predict(X)[0]
        return f"${price:,.2f}"

predictor = HousePricePredictor()

def create_price_map(min_price, max_price, sample_size):
    """Create interactive map of house prices"""
    df = pd.read_csv(DATA_PATH)

    # Filter by price range
    df_filtered = df[(df['price'] >= min_price) & (df['price'] <= max_price)]

    # Limit sample size for performance
    if len(df_filtered) > sample_size:
        df_filtered = df_filtered.sample(n=sample_size, random_state=42)

    # Create the map
    fig = px.scatter_mapbox(
        df_filtered,
        lat='lat',
        lon='long',
        color='price',
        size='sqft_living',
        hover_data={
            'price': ':$,.0f',
            'bedrooms': True,
            'bathrooms': True,
            'sqft_living': ':,',
            'waterfront': True,
            'lat': ':.4f',
            'long': ':.4f'
        },
        color_continuous_scale='Viridis',
        zoom=9,
        height=500,
        title=f'Seattle/King County House Prices ({len(df_filtered):,} houses shown)'
    )

    fig.update_layout(
        mapbox_style="open-street-map",
        margin={"r":0,"t":40,"l":0,"b":0}
    )

    return fig

def predict_price(bedrooms, bathrooms, sqft, age):
    """Make a price prediction"""
    return predictor.predict(bedrooms, bathrooms, sqft, age)

# Build Gradio Interface
with gr.Blocks() as demo:
    # Custom CSS for Helvetica Neue font
    gr.HTML("""
    <style>
    * {
        font-family: 'Helvetica Neue', Helvetica, Arial, sans-serif !important;
    }
    </style>
    """)

    gr.Markdown(
        """
        # 🏠 House Price Predictor
        ### AI-Powered Real Estate Valuation - Seattle/King County

        Trained on **21,613 real house sales** from King County, Washington (2014-2015).
        """
    )

    # Map Visualization Section
    gr.Markdown("## 🗺️ Explore House Prices on Map")

    with gr.Row():
        min_price_filter = gr.Slider(
            minimum=0,
            maximum=8000000,
            value=0,
            step=100000,
            label="Min Price ($)",
            info="Minimum price"
        )

        max_price_filter = gr.Slider(
            minimum=0,
            maximum=8000000,
            value=2000000,
            step=100000,
            label="Max Price ($)",
            info="Maximum price"
        )

        sample_slider = gr.Slider(
            minimum=100,
            maximum=2000,
            value=1000,
            step=100,
            label="Houses to Display",
            info="More = slower"
        )

        update_map_btn = gr.Button("🔄 Update Map", variant="secondary", size="sm")

    map_plot = gr.Plot(label="Interactive Price Map")

    # Load initial map
    demo.load(
        fn=lambda: create_price_map(0, 2000000, 1000),
        outputs=map_plot
    )

    # Update map on button click
    update_map_btn.click(
        fn=create_price_map,
        inputs=[min_price_filter, max_price_filter, sample_slider],
        outputs=map_plot
    )

    gr.Markdown("---")

    # Price Prediction Section
    gr.Markdown("## 🔮 Predict House Price")

    with gr.Row():
        with gr.Column():
            bedrooms = gr.Slider(1, 10, value=3, step=1, label="Bedrooms")
            bathrooms = gr.Slider(1, 10, value=2, step=1, label="Bathrooms")
            sqft = gr.Slider(500, 10000, value=2000, step=100, label="Square Feet")
            age = gr.Slider(0, 100, value=5, step=1, label="Age (years)")

            predict_btn = gr.Button("🔮 Predict Price", variant="primary", size="lg")

        with gr.Column():
            output = gr.Textbox(
                label="Predicted Price",
                placeholder="Click 'Predict Price' to see the result",
                scale=2
            )

            gr.Markdown(
                """
                ### Example Houses:
                - **Starter Home**: 2br, 1ba, 1200 sqft, 15 years
                - **Family Home**: 4br, 3ba, 2500 sqft, 10 years
                - **Luxury Home**: 5br, 4ba, 4000 sqft, 2 years
                """
            )

    gr.Examples(
        examples=[
            [2, 1, 1200, 15],  # Starter
            [3, 2, 2000, 5],   # Average
            [4, 3, 2500, 10],  # Family
            [5, 4, 4000, 2],   # Luxury
        ],
        inputs=[bedrooms, bathrooms, sqft, age],
        label="Quick Examples"
    )

    predict_btn.click(
        fn=predict_price,
        inputs=[bedrooms, bathrooms, sqft, age],
        outputs=output
    )

    gr.Markdown(
        """
        ---
        **Dataset:** 21,613 King County house sales (2014-2015)
        **Model:** Random Forest (R² Score: 0.53 on test data)
        **Architecture:** Gradio Frontend → ML Model (scikit-learn)
        """
    )

if __name__ == "__main__":
    demo.launch()