Enhancing Predictive Power in Financial Markets: Leveraging Autoencoders for Time Series Embeddings in Capital Markets

This project aims to develop a robust foundation model designed explicitly for financial time series representation learning. The core of this approach is an autoencoder framework capable of capturing multi-modal relationships in financial data. Once trained, the encoder will be deployed as a general-purpose model for various downstream financial tasks, including predictive analytics and asset pricing. A key focus is learning representations that enhance predictive accuracy and adhere to the strict no-arbitrage conditions in financial theory, ensuring theoretical consistency. This model is intended to bridge the gap between practical financial predictions and the foundational economic principles governing market behaviour, enabling its application across diverse financial contexts.

Faculty Supervisor:

Anastasis Kratsios

Student:

Partner:

Bank of Montreal

Discipline:

Mathematics

Sector:

Finance and Insurance

University:

McMaster University

Program: