Logistic Regression

• A supervised learning algorithm used for classification problems
• Sigmoid Function: values range from 0 - 1
• Assumptions
  1. Response Variable is Binary
  2. Independence - observations in the dataset are independent of each other
  3. No Multicollinearity - none of the predictor variables are highly correlated with each other
  4. No Extreme Outliers
  5. Linear Relationship between Explanatory Variables and Logit of Response Variable
  6. Sample Size is Sufficiently Large

• Multiclass Classification: One-vs-Rest (OvR)
• Approaches
  • Gradient Descent
    • Binary Cross Entropy / Log Loss
    • Cost(hθ(x),y) = −y log(hθ(x))−(1−y) log(1−hθ(x))
• Model Performance Evaluation
  • Confusion Matrix, Accuracy, Precision, Recall, F1-Score, ROC Curve