Non-linear classification refers to the process of categorizing data points that cannot be separated by a straight line, using algorithms that can model complex decision boundaries. This approach is crucial for handling real-world data that often exhibits intricate patterns and relationships, beyond the capabilities of linear classifiers.