Project: House Price Predictor
This is your capstone for tabular data. You will go from raw CSV to a deployed API in one project, using everything from earlier lessons: cleaning, splitting, pipeline, cross-validation, and random forest.
You will do exploratory data analysis first: plot price vs sqft, check missing values, and see correlations. Then build a pipeline that imputes, scales, and trains. Finally you will save and serve it.
The goal is not perfect accuracy but a working, explainable system you can demo to employers.
EDA to pipeline
Start with pandas describe and a scatter plot. Build a pipeline so preprocessing does not leak.
import pandas as pd
import matplotlib.pyplot as plt
df = pd.read_csv("housing.csv")
df.plot.scatter(x="sqft", y="price")
plt.show()Train and save
Train a forest, check cross-validation, save the pipeline.
from sklearn.ensemble import RandomForestRegressor
from sklearn.pipeline import Pipeline
pipe = Pipeline([("model", RandomForestRegressor(n_estimators=200, random_state=42))])
pipe.fit(X_train, y_train)
import joblib; joblib.dump(pipe, "house.joblib")Syntax
from sklearn.ensemble import RandomForestRegressor
model = RandomForestRegressor(n_estimators=200)
model.fit(X_train, y_train)
print(model.score(X_test, y_test))Data leakage in EDA
Looking at test data during EDA and then using those insights to choose features leaks test information. Do EDA only on train. This Kaggle post explains why leaderboard drops happen.
# Do EDA on train only: df_train = df.iloc[train_idx]Good to know
- EDA first, then pipeline.
- Use cross-validation, not single split.
- Save the whole pipeline.