app.py

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error, r2_score
import gradio as gr

# URL to the Excel dataset on Hugging Face
data_url = "https://huggingface.co/datasets/leadingbridge/flat/resolve/main/NorthPoint30.xlsx"

# Load dataset (using openpyxl as the engine)
df = pd.read_excel(data_url, engine="openpyxl")

# Drop columns that are not needed for prediction
cols_to_drop = ['Usage', 'Address', 'PricePerSquareFeet', 'InstrumentDate', 'Floor', 'Unit']
df.drop(columns=cols_to_drop, inplace=True, errors='ignore')

# Rename useful columns for consistency
df.rename(columns={"Floor.1": "Floor", "Unit.1": "Unit"}, inplace=True)

# Ensure the dataset has the necessary columns
required_columns = ['District', 'PriceInMillion', 'Longitude', 'Latitude', 'Floor', 'Unit', 'Area', 'Year', 'WeekNumber']
if not all(col in df.columns for col in required_columns):
    raise ValueError("Dataset is missing one or more required columns.")

# Define features and target variable
feature_names = ['District', 'Longitude', 'Latitude', 'Floor', 'Unit', 'Area', 'Year', 'WeekNumber']
X = df[feature_names]
y = df['PriceInMillion']

# Split into training and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Define a parameter grid for RandomForestRegressor and perform grid search
rf_param_grid = {
    'n_estimators': [50, 100, 150],
    "max_depth": [4, 6, 8],
    "max_features": ['sqrt', 'log2', 3],
    "random_state": [42]
}

rf_grid = GridSearchCV(RandomForestRegressor(), rf_param_grid, refit=True, verbose=1, cv=5, error_score='raise')
rf_grid.fit(X_train, y_train)

# Use the best estimator
model = rf_grid.best_estimator_

# Print model performance on the test set
rf_pred = model.predict(X_test)
rf_rmse = np.sqrt(mean_squared_error(y_test, rf_pred))
rf_r2 = r2_score(y_test, rf_pred)
print("Random Forest RMSE: ", rf_rmse)
print("Random Forest R2:   ", rf_r2)

def price_prediction(model, feature_names, new_data):
    new_data_df = pd.DataFrame([new_data], columns=feature_names)
    prediction = model.predict(new_data_df)
    return prediction[0]

def predict(district, longitude, latitude, floor, unit, area, year, weeknumber):
    new_data = [district, longitude, latitude, floor, unit, area, year, weeknumber]
    prediction = price_prediction(model, feature_names, new_data)
    return f"${prediction:,.2f} Million"

# Gradio Interface
iface = gr.Interface(
    fn=predict,
    inputs=[
         gr.Dropdown(choices=list(range(1, 9)), label='District (1 = Taikoo Shing, 2 = Mei Foo Sun Chuen, 3 = South Horizons, 4 = Whampoa Garden)'),
         gr.Number(label='Longitude'),
         gr.Number(label='Latitude'),
         gr.Dropdown(choices=list(range(1, 71)), label='Floor'),
         gr.Dropdown(choices=list(range(1, 31)), label='Unit (e.g., A=1, B=2, C=3, ...)'),
         gr.Slider(minimum=137, maximum=5000, step=1, label='Area (in sq. feet)'),
         gr.Dropdown(choices=[2024, 2025], label='Year'),
         gr.Dropdown(choices=list(range(1, 53)), label='Week Number')
    ],
    outputs=gr.Textbox(label='Price Prediction'),
    title="PROPERTY PRICE PREDICTION TOOL (Larry Pang Final Year Project)",
    description="Predict the price of a new property based on District, Longitude, Latitude, Floor, Unit, Area, Year, and Week Number."
)

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