Goal
In this music data science project, the overarching goal is to make music recommendations based on identified music taste.
Motivation
I spend a considerable amount of time relentlessly looking for new songs to fall in love with. While I very much enjoy doing that, I aim to make this process more efficient. Aside from the recommendations from Spotify via the personalized 'Discover Weekly' & 'Release Radar' playlists, I have been looking into the weekly featured new releases on Allmusic for many years. While the treasure hunt on Allmusic has always been enjoyable and fruitful, it is also very time intensive to go on the website, click on featured albums one by one, read the reviews, find the albums and songs on Spotify, listen to the songs one by one, and eventually find a handful of gems among a mountain of songs which I do not enjoy. Therefore, in this project, we build our personal AI music assistant to help us find music we will like faster.
Tools
Web scraping: BeautifulSoup, HTML, JSON
Databases: Google Cloud postgreSQL, IBM Cloud Db2, SQL, pandas
API: Spotify API, tekore
Test statistics: SciPy, statsmodels, Q-Q plot, Mann-Whitney U test, Shapiro-Wilk test, p-value
Machine learning: scikit-learn, NumPy, Hierarchical Clustering Analysis, Euclidian distance, PyTorch
Plots: Matplotlib, Plotly