This post is about some basic conceptions of model evaluation.

For predicting two-categry data, there are 4 possible results: True Positives, False Negatives, False Positives and True Negatives. To explain these conpects clearly, we can reference the picture below.

The blue dot represents the positive result while the red dot represents the negative result. Thus, the blue dots above the black line represent true positives, the blue dots below the line represents the false negatives, the red dots above the line represent the false positives and the red dots below the line represent the true negatives. The matrix in the picture above is the confusion matrix.

Accuracy

Accuracy = (True Positives + True Negatives) / Total numbers

Thus, the accuracy of the example above is 11/14=78.57%.

Regressioin Indicator

Mean Absolute Error

Mean Squared Error

R2 Error

Error Type

Underfitting

Overfitting

Tradeoff

Model Complexity Graph

Cross Validation