Course Summary: 

This module provides an applied introduction to modern forecasting and machine learning techniques for financial and macroeconomic prediction. It begins by exploring the principles of forecasting design and evaluation, including rolling-window approaches, multi-step forecasting, and overlapping forecast schemes. Students will learn to assess predictive accuracy using established statistical methods, such as the Diebold–Mariano test, alongside more advanced techniques, including Reality Checks, Superior Predictive Ability tests, and Model Confidence Sets. Particular emphasis is placed on the selection of appropriate loss functions, robustness analysis, and the reliability of forecast comparisons.

The module then examines forecasting in high-dimensional settings, where large numbers of potential predictors present challenges for model estimation and selection. Students will explore sparse modelling and regularisation techniques, including LASSO, Ridge Regression, and Elastic Net, to identify relevant predictors and improve predictive performance. The module also introduces non-linear forecasting approaches, including splines and Generalised Additive Models (GAMs), enabling students to capture complex relationships in financial and macroeconomic data. Students will subsequently investigate tree-based and ensemble learning methods, including regression and classification trees, bagging, random forests, and boosting algorithms. Cross-validation techniques for penalised regression and tree-based models are also covered, providing a rigorous framework for model selection, controlling overfitting, and evaluating out-of-sample predictive performance.

The module further explores unsupervised learning methods for extracting meaningful patterns from complex datasets. Students will study clustering techniques, including k-means and hierarchical clustering, to identify underlying structures in financial data, alongside dimension-reduction methods such as Principal Component Analysis (PCA), Principal Component Regression (PCR), and Partial Least Squares (PLS) to manage high-dimensional information and improve forecasting efficiency. Throughout the module, emphasis is placed on the critical evaluation and practical application of statistical and machine learning methods.

Topics: 

1. Forecasting Design & Evaluation Methods

2. Introduction to Penalized Regressions

3. Hyperparameter Selection and Market Timing

4. From Linear to Non-linear Methods

5. Tree-based Methods

6. Unsupervised Learning