Classification

Decision Tree Model:

Accuracy: 82,2%

Cohen's kappa: 50%

Random Forest Model:

Accuracy: 87%

Cohen's kappa: 61,3%

Used for Feature Selection. Importance of features calculated by # of splits they were used in (Information gain of the features)

Best performing Model

Tried optimizing with more Models (200 instead of 100): Only lead to a slight increase of Cohen's kappa to 61.5%, Accuracy stayed the same

Logistic Regression Model:

Accuracy: 85,2%

Cohen's kappa: 56,5%

Removed the 2 features with by far the least importance. Accuracy stayed the same. Removing the 3rd least important feature (hour per week) reduced the Accuracy by 0.2% to 85%

kNN Model: k=3, 5, 7

Accuracy: 82.3, 82.8, 83.1%

Cohen's kappa: 48, 49.3, 50%

Used One-Hot Encoding to convert strings to ints.