Classification

Predicts categories (numeric or non numeric)

Classification: Breast cancer detection

• One input
  • We only predict a finite set of outputs/categories/classes (in this case, 0 or 1)
  • If I have a data in that ?, would it be benign (0) or malignant (1)?
• 2 or more inputs
  • You can use more than 1 input value to predict the output
  • Example: Age and tumor size
  • The learning algorithm might find some boundary that separates the benign and malignant → it has to decide how to fit a boundary line through this data
  • So in out case with the yellow patient, the tumor is likely to be benign

Classification algorithms

• Logistic regression
• Decision tree/random forest
• neural networks
• naive bayes
• k-nearest neighbors
• support vector machines