Its Time to Evaluate!

07 Feb 2021

Hi there!

Lets discuss various ways to Evaluate an ML Model.

Basic Terminology:

• True Positive(TP): Predictive Positive is actually Positive.
• False Positive(FP): Predictive Positive is actually Negetive
• True Negetive(TN): Predective Negetive is actually Negetive
• False Negetive(FN): Predictive Negetive is actually Negetive

Confusion Matrix:

Accuracy:

Accuracy = (TP + TN)/(TP + TN + FP + FN)

Precision:

How much the model is Right when it says it is Right!

Precision = (TP)/(TP + TN)

Total Positive Rate/Recall/Senstivity:

% of Positive Instances out of Total Positive Instances

TPR = (TP)/(TP + FN)

Specificity:

% of Negetive Instances out of Total Negetive Instances

Specificity = (TN)/(TN + FP)

F1 Score:

It is the Harmonic Mean of Precision and Recall. The higher the F1 Score, the better it is.

PR Curve:

It is the curve between precision and recall for various threshold values.

Lets catch-up in the next post.

Till then O/