You are an ML engineer at a global car manufacture. You need to build an ML model to predict car sales in different cities around the world. Which features or feature crosses should you use to train…

Question

Options

• AThee individual features: binned latitude, binned longitude, and one-hot encoded car type.
• BOne feature obtained as an element-wise product between latitude, longitude, and car type.
• COne feature obtained as an element-wise product between binned latitude, binned longitude, and
• DTwo feature crosses as an element-wise product: the first between binned latitude and one-hot

• A
  2% (1)
• B
  5% (2)
• C
  82% (36)
• D
  11% (5)

Explanation

The goal is to capture city-specific relationships between car type and sales volume. A 'city' is defined by both latitude and longitude together, so both dimensions must be combined. Binning continuous lat/lon values discretizes the space into geographic regions (city-like buckets). Creating a single three-way feature cross of binned_latitude × binned_longitude × one-hot_car_type captures the joint interaction: 'in this specific city, this specific car type has this sales pattern.' Option A uses three separate features, which cannot capture the multiplicative interaction between location and car type. Option B does not bin lat/lon, so continuous values produce a poor feature cross that doesn't discretize into meaningful city regions. Option D creates two separate two-way crosses (binned_lat × car_type and binned_lon × car_type), but these cannot represent city-level specificity because a city requires both lat and lon to be identified simultaneously-splitting them loses the joint geographic identity.

No community discussion yet for this question.