The project i got to pursue in databases was by far the closest to my heart.
Since music is such a big part of my life, getting the opportunity to design a database for anything i choose was the best kind of freedom given to me. I went straight to Spotify, Soundcloud, iTunes, Youtube and began reading about how they design their music recommendation algorithms.
I proposed to several people my idea of incorporating one into a database itself so as to save on processing power or time. I was told by many that it would be too complicated to do for a project which lasts just one semester. I persevered.
I struggled with figuring out how to join my tables, what types of joins to use and how to design a query which would work like a recommendation algorithm. I came up with a solution which i compartmentalized to ensure execution. I worked up a mood playlist query which would take user input and generate a playlist containing songs which pertain to their mood. Attaching certain parameters to each mood setting to ensure the genre, tempo and other parameters of the music to match with the mood.
After that i designed a query which took a song based off of which a recommendation would be made and within certain ranges of BPM, Tempo, Octave the other matching song would be recommended and the count of recommendations was up to the user. This was not remotely as glamorous a recommendation algorithm as Googles Deepmind, but nonetheless i was able to prove that a recommendation algorithm of sorts can be incorporated into a database query itself and was successful at creating forms, views and interactive reports and gave a successful demonstration too.