Intended for

Empirical researchers interested in using high-dimensional or unstructured data in their projects.

Prerequisites

Knowledge of probability theory equivalent to advanced undergraduate or graduate courses in econometrics. The instructor will circulate voluntary programming exercises that can be done in any scripting language; students will not be required to complete these but they will be discussed in class. Live demonstrations will be in Python.

Overview

This course shows how to apply modern statistical techniques to big financial data. The focus is on how machine learning can guide academic research in Finance, as well as decisions in the financial industry, including asset managers, hedge funds, and consumer finance companies. We will cover unsupervised and supervised machine learning techniques and their applications in asset pricing and credit scoring. We will also cover reinforcement learning, with applications to portfolio choice. The primary purpose of this course is not only to teach statistical methods, but also to facilitate the financial and economic interpretation of machine learning. Hence, we will pay special attention to the interpretability of machine-learning results, and to the distinction between correlation and causation.

Topics

Data-driven asset management
Supervised and unsupervised machine learning models
Reinforcement learning and portfolio choice
Consumer credit markets
Interpretability and causal inference in machine learning

Ansgar Walther is Assistant Professor of Finance at Imperial College London. His research and teaching focuses on financial economics, the economics of information, machine learning and FinTech. He obtained his PhD in Economics from the University of Cambridge, and has held academic posts at Oxford and Warwick.