Exploring House Price Data

import pandas as pd
import numpy as np

# Build the NestView listings dataset
df = pd.DataFrame({
    "property_id":   ["P001","P002","P003","P004","P005","P006",
                      "P007","P008","P009","P010","P011","P012"],
    "neighbourhood": ["Riverside","Oldtown","Riverside","Greenhill","Oldtown","Greenhill",
                      "Riverside","Greenhill","Oldtown","Riverside","Greenhill","Oldtown"],
    "price":         [420000,285000,565000,340000,210000,375000,
                      490000,310000,255000,620000,395000,230000],
    "sqft":          [1850,1420,2300,1650,1100,1580,
                      2050,1720,1280,2480,1900,1200],
    "bedrooms":      [3,3,4,3,2,3,4,3,2,4,3,2],
    "age_years":     [12,45,8,22,60,18,5,15,52,3,10,58],
    "condition":     [8,5,9,7,4,7,9,8,5,10,8,4]
})

# Inspect
print("Shape:", df.shape)
print("Missing values:", df.isnull().sum().sum())

# Price distribution
price_stats = df["price"].describe()
print("\nPrice distribution:")
print(price_stats.round(0))

# Mean vs median — how skewed is the distribution?
mean_price   = df["price"].mean()
median_price = df["price"].median()
skew         = (mean_price - median_price) / df["price"].std()

print(f"\nMean price:   ${mean_price:,.0f}")
print(f"Median price: ${median_price:,.0f}")
print(f"Skew index:   {skew:.3f} ({'right-skewed' if skew > 0 else 'left-skewed'})")
print(f"Price range:  ${df['price'].min():,} to ${df['price'].max():,}")

# Compute price per square foot — the normalised value metric
df["price_per_sqft"] = (df["price"] / df["sqft"]).round(2)

print("Price per sqft for all listings:")
print(df[["property_id","neighbourhood","price",
          "sqft","price_per_sqft"]].sort_values(
    "price_per_sqft", ascending=False
).to_string(index=False))

# Group by neighbourhood — mean price, mean price per sqft, listing count
neighbourhood = df.groupby("neighbourhood").agg(
    listings        = ("property_id",   "count"),
    mean_price      = ("price",         "mean"),
    mean_price_sqft = ("price_per_sqft","mean"),
    mean_sqft       = ("sqft",          "mean"),
    mean_condition  = ("condition",      "mean")
).round(1).reset_index()

neighbourhood = neighbourhood.sort_values(
    "mean_price_sqft", ascending=False
)

print("\nNeighbourhood performance summary:")
print(neighbourhood.to_string(index=False))

# Location premium: gap between most and least expensive per sqft
top    = neighbourhood.iloc[0]
bottom = neighbourhood.iloc[-1]
premium_pct = (top["mean_price_sqft"] - bottom["mean_price_sqft"]) / bottom["mean_price_sqft"] * 100
print(f"\nLocation premium ({top['neighbourhood']} vs {bottom['neighbourhood']}): {premium_pct:.1f}%")

# Compute Pearson correlation between price and all numeric features
features = ["sqft","bedrooms","age_years","condition","price_per_sqft"]
print("Correlation with price (Pearson r):")
for feat in features:
    r = df["price"].corr(df[feat]).round(3)
    bar = "#" * int(abs(r) * 20)
    direction = "+" if r > 0 else "-"
    print(f"  {feat:<18} r = {r:+.3f}  {bar}")

# Full correlation matrix for all numeric columns
numeric_cols = ["price","sqft","bedrooms","age_years","condition","price_per_sqft"]
corr_matrix  = df[numeric_cols].corr().round(3)
print("\nFull correlation matrix:")
print(corr_matrix.to_string())

# Does condition compensate for age?
age_cond_corr = df["age_years"].corr(df["condition"]).round(3)
print(f"\nAge vs condition correlation: {age_cond_corr:.3f}")
print("Interpretation: older properties tend to have", 
      "worse" if age_cond_corr < 0 else "better", "condition scores")

# Compute neighbourhood-level mean and std of price_per_sqft
# We will use these to flag outliers within each neighbourhood
neighbourhood_stats = df.groupby("neighbourhood")["price_per_sqft"].agg(
    mean_psf = "mean",
    std_psf  = "std"
).reset_index()

# Merge neighbourhood stats back onto the main DataFrame
df = df.merge(neighbourhood_stats, on="neighbourhood")

# Compute z-score: how many std deviations each property is from its
# neighbourhood mean price per sqft
df["psf_zscore"] = (
    (df["price_per_sqft"] - df["mean_psf"]) / df["std_psf"]
).round(2)

# Flag undervalued properties: z-score below -1.0
# (more than 1 std below the neighbourhood mean price per sqft)
undervalued = df[df["psf_zscore"] < -1.0][
    ["property_id","neighbourhood","price","sqft",
     "price_per_sqft","mean_psf","psf_zscore","condition"]
].sort_values("psf_zscore")

print("Undervalued properties (psf z-score < -1.0):")
print(undervalued.to_string(index=False))

# Show full z-score table for context
print("\nAll properties with neighbourhood z-scores:")
print(df[["property_id","neighbourhood","price_per_sqft",
          "mean_psf","psf_zscore"]].sort_values(
    ["neighbourhood","psf_zscore"]
).to_string(index=False))