Cohen Kappa

INFO

Measures agreement between predicted and actual classifications while accounting for chance agreement.

How It Works

Cohen’s Kappa compares observed accuracy with expected accuracy (random chance):

$$\kappa =\frac{p_o-p_e}{1-p_e}$$

• $p_o$: Observed agreement
• $p_e$: Expected agreement by chance

What to Look For

• Values range from -1 to 1
• ($\kappa >0.6$): Substantial agreement
• Useful for multi-class classification and inter-rater reliability

Application Models

• Random Forest Classification
• Support Vector Machine (SVM)
• Neural Network Regression