Google Professional Data Engineer
Get started today
Ultimate access to all questions.
Ultimate access to all questions.
You are working on a project to train a model aimed at predicting housing prices using a provided dataset of real estate properties. The dataset includes various features, one of which is the latitude and longitude of each property. Given that real estate experts have emphasized the significant impact of a property's location on its market price, you have decided to integrate this geographic information into your predictive model. You plan to train a fully connected neural network. Considering the critical nature of location in determining real estate prices, what approach should you take to effectively engineer a feature that captures this geographic dependency?
Explanation:
The correct answer is C. Using a feature cross of latitude and longitude helps capture the spatial relationship between the two geographical features, which is crucial for predicting housing prices. Bucketizing the feature cross at the minute level can capture localized patterns. L1 regularization is preferable in this case because it assigns greater importance to more influential features and shrinks less important features to zero, which is useful for features like location that are known to be influential in predicting housing prices.