Microsoft Azure AI Fundamentals (AI-900) Practice Exam 2026 - Free AI-900 Practice Questions and Study Guide

Evaluating a machine learning model using the same data that was used for training is not advisable because it can lead to misleading results. This practice is known as "overfitting." When a model is evaluated on the training dataset, it typically performs very well, as it has learned the patterns, noise, and specificities of that data. However, this performance can mask the model's true ability to generalize to new, unseen data.

The main goal of a machine learning model is to accurately predict or classify data that it has not encountered before. To achieve this, the model should be evaluated on a separate dataset that it hasn't seen during the training process, often referred to as a validation or test dataset. This separation allows for a more accurate assessment of the model's performance in real-world scenarios, ensuring that it can generalize beyond the training data.

Through this process, you can determine metrics such as accuracy, precision, recall, and F1 score in a way that reflects how the model will perform in practice, rather than just on the training examples it has memorized.