Mini Project: EDA Report on a Real Dataset

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.datasets import fetch_california_housing

# LOAD DATA
housing = fetch_california_housing()
df = pd.DataFrame(housing.data, columns=housing.feature_names)
df["MedHouseVal"] = housing.target  # median house value in $100k

print("=== CALIFORNIA HOUSING EDA REPORT ===")
print(f"\nDataset: {df.shape[0]} census blocks, {df.shape[1]} features")
print("\n1. DATA OVERVIEW")
print(df.describe().round(2))

print("\n2. MISSING VALUES:", df.isnull().sum().sum(), "(none expected in sklearn datasets)")

# KEY FEATURE DESCRIPTIONS
feature_desc = {
    "MedInc":     "Median income (in $10,000s)",
    "HouseAge":   "Median house age (years)",
    "AveRooms":   "Average rooms per household",
    "AveBedrms":  "Average bedrooms per household",
    "Population": "Block population",
    "AveOccup":   "Average household occupants",
    "Latitude":   "Geographic latitude",
    "Longitude":  "Geographic longitude",
    "MedHouseVal":"TARGET: Median house value (in $100,000s)",
}
print("\n3. FEATURE DESCRIPTIONS:")
for feat, desc in feature_desc.items():
    print(f"  {feat:12s}: {desc}")

# DISTRIBUTION ANALYSIS
fig, axes = plt.subplots(3, 3, figsize=(15, 12))
axes = axes.ravel()

for i, col in enumerate(df.columns):
    axes[i].hist(df[col], bins=40, color="steelblue", edgecolor="white", alpha=0.8)
    axes[i].set_title(col)
    skew = df[col].skew()
    axes[i].set_xlabel(f"skew={skew:.2f}" + (" (log needed)" if abs(skew) > 1 else ""))

plt.suptitle("Feature Distributions — California Housing", y=1.02, fontsize=14)
plt.tight_layout()
plt.savefig("housing_distributions.png", dpi=100, bbox_inches="tight")
plt.show()

# CORRELATION WITH TARGET
print("\n4. CORRELATION WITH TARGET (MedHouseVal):")
corr_target = df.corr()["MedHouseVal"].drop("MedHouseVal").sort_values(ascending=False)
for feat, corr in corr_target.items():
    bar = "+" * int(abs(corr) * 20) if corr > 0 else "-" * int(abs(corr) * 20)
    print(f"  {feat:12s}: {corr:+.3f} {bar}")

# GEOGRAPHIC VISUALIZATION
fig, ax = plt.subplots(figsize=(10, 7))
scatter = ax.scatter(
    df["Longitude"], df["Latitude"],
    c=df["MedHouseVal"], cmap="RdYlGn",
    s=df["Population"] / 200, alpha=0.5,
)
plt.colorbar(scatter, ax=ax, label="Median House Value ($100k)")
ax.set_xlabel("Longitude")
ax.set_ylabel("Latitude")
ax.set_title("California Housing: Price & Population by Location\n(dot size = population)")
plt.tight_layout()
plt.savefig("housing_map.png", dpi=100, bbox_inches="tight")
plt.show()

# EDA CONCLUSIONS -- what to do in preprocessing
print("\n5. EDA CONCLUSIONS (for preprocessing):")
conclusions = [
    "MedInc: highest correlation (+0.69) with target -- most important feature",
    "Population and AveOccup: right-skewed, consider log transform",
    "AveRooms/AveBedrms: some very large values (outliers -- gated communities?)",
    "Latitude/Longitude: geographic features, consider interaction term",
    "No missing values -- skip imputation",
    "All features numeric -- no encoding needed",
    "Target capped at 5.0 (500k) -- be aware of truncation",
]
for c in conclusions:
    print(f"  -> {c}")