Mini Project: House Price Prediction

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import train_test_split, cross_val_score, GridSearchCV
from sklearn.preprocessing import StandardScaler, FunctionTransformer
from sklearn.pipeline import Pipeline
from sklearn.ensemble import GradientBoostingRegressor, RandomForestRegressor
from sklearn.linear_model import Ridge
from sklearn.metrics import mean_absolute_error, r2_score
import joblib

# ━━━━━━━━━━━━━━━━━━━━━━━━━━
# STEP 1: LOAD & FEATURE ENGINEERING
# ━━━━━━━━━━━━━━━━━━━━━━━━━━
housing = fetch_california_housing()
df = pd.DataFrame(housing.data, columns=housing.feature_names)
df["MedHouseVal"] = housing.target

# EDA-driven feature engineering
df["rooms_per_person"] = df["AveRooms"] / df["AveOccup"]
df["bedrooms_ratio"]   = df["AveBedrms"] / df["AveRooms"]
df["income_per_room"]  = df["MedInc"] / (df["AveRooms"] + 1)
# Clip to remove unrealistic values
df["rooms_per_person"] = df["rooms_per_person"].clip(0, 15)
df["bedrooms_ratio"]   = df["bedrooms_ratio"].clip(0, 1)

feature_cols = housing.feature_names.tolist() + ["rooms_per_person", "bedrooms_ratio", "income_per_room"]
X = df[feature_cols]
y = df["MedHouseVal"]

print(f"Features: {feature_cols}")
print(f"Shape: {X.shape}")

# ━━━━━━━━━━━━━━━━━━━━━━━━━━
# STEP 2: SPLIT
# ━━━━━━━━━━━━━━━━━━━━━━━━━━
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.15, random_state=42)
X_train, X_val,  y_train, y_val  = train_test_split(X_train, y_train, test_size=0.15, random_state=42)
print(f"Train: {len(X_train)} | Val: {len(X_val)} | Test: {len(X_test)}")

# ━━━━━━━━━━━━━━━━━━━━━━━━━━
# STEP 3: MODEL COMPARISON
# ━━━━━━━━━━━━━━━━━━━━━━━━━━
models = {
    "Ridge":            Pipeline([("scaler", StandardScaler()), ("model", Ridge(alpha=10.0))]),
    "Random Forest":    Pipeline([("model", RandomForestRegressor(n_estimators=100, random_state=42, n_jobs=-1))]),
    "Gradient Boost":   Pipeline([("model", GradientBoostingRegressor(n_estimators=200, learning_rate=0.1, max_depth=5, random_state=42))]),
}

print("\nModel comparison (5-fold CV on training data):")
for name, model in models.items():
    cv_r2 = cross_val_score(model, X_train, y_train, cv=5, scoring="r2")
    print(f"  {name:<18}: CV R2 = {cv_r2.mean():.4f} +/- {cv_r2.std():.4f}")

# ━━━━━━━━━━━━━━━━━━━━━━━━━━
# STEP 4: TUNE BEST MODEL
# ━━━━━━━━━━━━━━━━━━━━━━━━━━
param_grid = {
    "model__n_estimators":  [200, 300],
    "model__learning_rate": [0.05, 0.1],
    "model__max_depth":     [4, 5, 6],
}
gb_pipeline = Pipeline([("model", GradientBoostingRegressor(random_state=42))])
search = GridSearchCV(gb_pipeline, param_grid, cv=5, scoring="r2", n_jobs=-1, verbose=0)
search.fit(X_train, y_train)
print(f"\nBest params: {search.best_params_}")
print(f"Best CV R2:  {search.best_score_:.4f}")

# ━━━━━━━━━━━━━━━━━━━━━━━━━━
# STEP 5: FINAL EVALUATION
# ━━━━━━━━━━━━━━━━━━━━━━━━━━
best_model = search.best_estimator_
best_model.fit(X_train, y_train)

val_r2  = best_model.score(X_val, y_val)
test_r2 = best_model.score(X_test, y_test)
test_mae = mean_absolute_error(y_test, best_model.predict(X_test))

print(f"\nFinal Results:")
print(f"  Validation R2: {val_r2:.4f}")
print(f"  Test R2:       {test_r2:.4f}  (honest estimate)")
print(f"  Test MAE:      ${test_mae * 100_000:,.0f}  (average prediction error)")

# SAVE MODEL
joblib.dump(best_model, "house_price_model.joblib")
print("\nModel saved to house_price_model.joblib")

# PREDICTION FUNCTION
def predict_house_price(rooms: float, income: float, lat: float, lon: float) -> str:
    sample = pd.DataFrame([{
        "MedInc": income, "HouseAge": 30, "AveRooms": rooms,
        "AveBedrms": rooms * 0.3, "Population": 1500, "AveOccup": 2.5,
        "Latitude": lat, "Longitude": lon,
        "rooms_per_person": rooms / 2.5, "bedrooms_ratio": 0.3,
        "income_per_room": income / (rooms + 1),
    }])[feature_cols]
    pred = best_model.predict(sample)[0]
    return f"${pred * 100_000:,.0f}"

print("\nSample predictions:")
print(f"  LA: {predict_house_price(5, 8, 34.05, -118.25)}")
print(f"  SF: {predict_house_price(4, 10, 37.77, -122.43)}")