Logistic Regression - Classification Problem

Classification

Examples:

• Email: Spam / Not Spam
• Online Transactions: Fraudulent (Yes / No)
• Tumor: Malignant / Benign

Above are 2 classes (binary) classification problems where $y \in {0, 1}$, 0 is negative class and 1 is positive class.

Multi classification problems: $y \in {0, 1, 2, 3, ...}$

Linear regression doesn't work as we added a sample with very big tumor size:

We also notice that linear regression will output values h(x) > 1 or < 0. This is more than what we need (0 or 1). In general, linear regression doesn't work for classification problems. With logistic regression: $0 \le h(x) \le 1$