Which assertion about independent variable data types in neural networks is correct?

• A. Only categorical data can be used
• B. Numeric data type is required for independent variables
• C. Text data types are preferred for modeling
• D. Any data type may be used for independent variables

Answer: (B)

Neural networks operate through arithmetic on inputs, weights, and activations. Everything in those computations must be numeric, because matrix multiplications, additions, and activation functions only accept numbers. If a feature isn’t numeric, you must transform it into a numeric representation before feeding it into the model (for example, one-hot encoding for categories or embeddings for text). That’s why the statement that a numeric data type is required for independent variables is the correct one: it captures the essential requirement that the network’s inputs be numeric, with non-numeric features converted into numbers prior to modeling.

The other ideas don’t fit because raw categories or text aren’t used directly by neural nets; they must be encoded into numeric form first, and saying any data type can be used as-is ignores this encoding step. Similarly, focusing on text as the preferred data type is misleading since it also requires numeric representation before the network can process it.