Design features for house price prediction

Scenario

You are building a model to predict house sale price from a tabular dataset (similar to typical real-estate datasets). The interviewer expects a simple baseline model (e.g., linear regression), but wants to understand your reasoning.

Questions

1. Which features are likely to be predictive of house price, and why? (Examples: location, size, age, condition, amenities, nearby schools, etc.)
2. How do you decide which features are usable (available at prediction time, not leaking label information, stable definitions)?
3. What data cleaning steps would you perform before modeling?
4. If starting with linear regression , how would you:
   • handle missing values,
   • handle categorical variables,
   • reduce the impact of outliers/skewed price distributions,
   • detect multicollinearity and mitigate it?
5. How would you evaluate the model and iterate on improvements?

Assume you have a training set with historical sales and a holdout set for evaluation.