app.py

import streamlit as st
import pandas as pd
import joblib

# Load saved model and encoder
artifacts = joblib.load("melb_model_artifacts.pkl")
model = artifacts["model"]
encoder = artifacts["encoder"]
feature_names = artifacts["feature_names"]

# Streamlit UI
st.title("🏠 Melbourne House Price Predictor")

st.markdown("Enter property details to get a price prediction:")

# Basic numeric inputs
rooms = st.number_input("Rooms", min_value=1, max_value=10, value=3)
bathroom = st.number_input("Bathroom", min_value=1, max_value=10, value=2)
car = st.number_input("Car Spaces", min_value=0, max_value=10, value=1)
landsize = st.number_input("Land Size (m²)", min_value=0.0, value=200.0)
building_area = st.number_input("Building Area (m²)", min_value=0.0, value=120.0)

# Categorical inputs — values must exist in training data
suburb = st.text_input("Suburb (e.g. 'Kew', 'Richmond', or 'Other')", "Other")
type_ = st.selectbox("Type", ["h", "u", "t"])
method = st.selectbox("Method", ["S", "SP", "PI", "VB"])
sellerg = st.text_input("SellerG (e.g. 'Nelson', 'Other')", "Other")
council = st.text_input("Council Area", "Other")
region = st.text_input("Region Name", "Other")

# Predict button
if st.button("Predict Price"):
    # Build a single-row dataframe from inputs
    data = {
        "Rooms": rooms,
        "Bathroom": bathroom,
        "Car": car,
        "Landsize": landsize,
        "BuildingArea": building_area,
        "Suburb": suburb,
        "Type": type_,
        "Method": method,
        "SellerG": sellerg,
        "CouncilArea": council,
        "Regionname": region
    }
    df = pd.DataFrame([data])

    # Match training preprocessing
    useful_cats = ['Suburb', 'Type', 'Method', 'SellerG', 'CouncilArea', 'Regionname']
    cats_df = df[useful_cats].fillna("Other")
    encoded = encoder.transform(cats_df)
    encoded_df = pd.DataFrame(encoded, columns=encoder.get_feature_names_out(useful_cats))
    numeric_df = df.select_dtypes(include=["int64", "float64"])
    X = pd.concat([numeric_df, encoded_df], axis=1).reindex(columns=feature_names, fill_value=0)

    # Predict
    prediction = model.predict(X)[0]
    st.success(f"Estimated Price: ${prediction:,.0f}")