Microsoft Azure AI Fundamentals (AI-900) Practice Exam

In a classification model, are labels required to be numeric?

1. Yes

2. No

3. Only for binary classification

4. Only for multiclass classification

The correct answer is: No

In a classification model, labels do not need to be numeric as they can also be categorical. Classification models are designed to sort data into predefined classes or categories, and these labels can be represented as strings or any other categorical types, not just numbers. For example, in a model trained to classify types of animals, the labels could be "cat," "dog," and "bird." These categorical labels provide meaningful identification of classes to the model. During preprocessing, if numerical representations are necessary for algorithms or libraries being used—especially those that expect numeric input—these categorical labels can be encoded into numeric formats. However, this encoding is a step of data preprocessing, not a requirement of the classification model itself. Thus, the core function of classification is to allocate instances to classes based on their features, and this can effectively be done with non-numeric labels too.