For many, music is a part of everyday life and can be linked to events they experience; it is linked to emotion and often used to improve our mood. This project aims to investigate how music listening habits were affected by the COVID-19 pandemic in 2020 – this will be achieved by researching if a sentiment and mood can be accurately obtained from a song’s lyrics among other song data provided by Spotify, the world’s largest music streaming platform. To approach this, the project will use natural language processing (specifically sentiment analysis) and machine learning to analyse the most popular songs within this timeframe to predict each song’s mood. Similarly, the most popular songs prior to COVID-19 and in the years after the initial outbreak will be examined to see if a change in listening trends did occur. While the classification of moods in songs is not a new concept, this project aims to attempt a new approach of testing the best machine learning model for it and using sentiment analysis of lyrics to create a richer dataset. Based on the evaluations, logistic regression was the most accurate model; the results also showed that there was a noticeable change in the number of songs per mood across each year. Additionally, the findings indicated that the model could predict moods accurately, but not consistently enough to be reliable. Limitations included the dataset size and analysis of only four years – further work using bigger datasets and a longer timeframe would build upon the project’s findings. In other applications, this research could be used to inspire the work of more personalised and unique song recommendations, or the creation of custom playlists used to improve one’s wellbeing, for example.