Machine Learning Guide for Oil and Gas Using Python

Metrics for classification model evaluation

A classification accuracy is defined as follows:

(5.9)

Although accuracy provides a general understanding of the model's accuracy in terms of predicting the correct classes, one major problem with using accuracy when evaluating a classification model is when there is unbalanced class distribution within a data set. For example, if in a data set with 1 MM rows, 100K rows are assigned to class 0 and the remaining 900K rows are assigned to class 1, this is classified as an unbalanced distribution. Therefore, confusion matrix is recommended to obtain more information in terms of evaluating a model's accuracy. ...