(HL) Data preprocessing and supervised learning

Learning objectives

(HL) Describe how data cleaning handles outliers, duplicates, incorrect, irrelevant and badly formatted values and missing values (by imputation, deletion or predictive modelling), and why features are normalized or standardized before training

(HL) Describe filter, wrapper and embedded methods of feature selection, and explain how the curse of dimensionality makes dimensionality reduction worthwhile

(HL) Explain how linear regression predicts a continuous outcome from independent variables, interpreting the slope, the intercept and the coefficient of determination R²

(HL) Explain how k-nearest neighbours and decision trees assign new data to discrete categories, with applications such as recommendation systems and diagnosis from symptoms

(HL) Evaluate a supervised model with accuracy, precision, recall and F1 score, and explain how hyperparameter tuning is used to avoid overfitting and underfitting

(HL) Explain why several models are compared before one is selected, justifying the choice by the nature of the data, the complexity of the problem and the outcome required