In a neural network, the dependent variable can be which data types?

• A. Numeric and categorical
• B. Only numeric
• C. Only categorical
• D. Text

Answer: (A)

Neural networks used for supervised learning can have dependent variables that are either continuous values or discrete categories. A numeric target corresponds to regression tasks where the network outputs a real number and is trained with a loss like mean squared error. A categorical target corresponds to classification tasks where the network outputs probabilities over classes via a softmax (or a sigmoid for binary) and is trained with cross-entropy loss. Text as a direct dependent variable isn’t typical in standard supervised learning because text must be converted into numbers (tokens, embeddings) before the network can process it, and models that generate text treat the problem as sequence generation rather than predicting a single numeric label. Therefore, the dependent variable can be numeric and categorical.