Design a house-price prediction model
Overview
This question evaluates machine learning and data science competencies including regression model design, feature engineering, data-splitting and leakage prevention, selection of evaluation metrics, handling missing values, outliers and high-cardinality location features, temporal drift management, and model interpretation for house-price prediction. It is commonly asked in the Machine Learning domain to assess end-to-end practical application and conceptual understanding of validation and metric trade-offs, primarily testing practical application supported by conceptual reasoning and stakeholder communication.
Problem
You are asked to build a model to predict house sale prices for a city of your choice.
Data (assume typical real-estate fields)
You have a historical dataset of home listings/sales with (examples):
-
sale_id(string, unique) -
city(string) -
sale_date(date) -
sale_price(float, target) -
bedrooms(int),bathrooms(float),sqft(float),lot_sqft(float) -
year_built(int) -
zipcode/neighborhood(string) -
lat/lon(float) -
property_type(categorical) -
days_on_market(int) - Optional external features (if you choose): school ratings, crime, interest rates, nearby transit, etc.
Tasks
- Propose an end-to-end approach (data cleaning → feature engineering → model selection).
- Define how you would split data (train/validation/test) and avoid leakage .
- Choose evaluation metrics and justify them (e.g., MAE vs RMSE vs MAPE).
-
Explain how you would handle:
- missing values and outliers
- high-cardinality location features (zip/neighborhood)
- temporal drift (market changes)
- Describe how you would interpret the model and communicate results to stakeholders.
Solution
Login required