Performance and error assessment measures

The metrics we use to assess the performance and calculate errors in our models, and how we interpret their values, determine the models we select, the decisions we make to improve a component of our machine learning life cycle, and determine if we have a reliable model to bring into production. Although many performance metrics can be used in one line of Python code to calculate errors and performance, we shouldn’t blindly use them or try to improve our performance reports by implementing many of them together without knowing their limitations and how to correctly interpret them. In this section, we will talk about metrics for assessing the performance of classification, regression, and clustering models.

Classification

Each classification model, either binary or multi-class, returns the probability of predictions, a number between 0 and 1, which then gets transformed into class labels. There are two major categories of performance...