A data scientist is building a regression model to predict house prices. The dataset contains a feature 'neighborhood' with 500 distinct values, and most neighborhoods have fewer than 10 samples. Which approach is MOST appropriate for handling this high-cardinality categorical feature?