Evaluation Metrics

In this section, we will go through all the evaluation measures for assessing the quality of the machine learning model predictions. Based on the dependent variable, we have several choices for the evaluation measures. In the train and evaluate step of our Machine Learning Workflow, we mentioned that until we get the desired results, we keep iterating the training model by adding new variables or changing the parameters. In each iteration, we try to optimize for any one or two evaluation metrics. The following table summarizes the various types of metrics used for regression, classification, and recommender systems. Given the scope of this book, we will delve into more details on regression and classification algorithms:

Figure 3.14: Metrics for various types of machine learning algorithms.

Mean Absolute Error (MAE)

Absolute error is direction-agnostic, which means that it is does not matter whether the predicted value of the dependent variable by the model on the test dataset...