Evaluating the Performance of the MOMENT Time-Series Foundation Model in Financial Markets: A Comparative Study with Traditional Models in the Stock Market

The stock market’s inherent volatility and complexity make it a challenging domain for traditional machine learning models. While advances in foundation models such as MOMENT, trained on large and diverse time-series datasets, have demonstrated strong performance across a range of tasks, their application to financial markets remains largely unexplored. This thesis aims to examine the stock market performance of the MOMENT model, by comparing it against scientific work which used traditional machine learning models across the key tasks, forecasting, classification, anomaly detection, and imputation. Beyond the comparative analysis, this work also aims to identify and address potential weaknesses of MOMENT, thereby making a unique contribution to the field.