Week 1 — Topify
Our aim is to predict the possibility of a song if it will be on Spotify’s top list.
As for why we chose this problem, being able to predict whether or not anything would be famous ahead of time is a challenging problem in any industry. It has also recently become a very significant topic in the dynamic music industry. Because of the widespread use of music apps (Spotify, SoundCloud, Deezer), data can be easily accessed and people’s listening habits can be easily monitored.
For this purpose, we are interested in what makes a song popular. To elaborate, we are looking for which attributes of a song impact popularity more. And then we will use this information for our prediction.
We will collect our data from mostly Spotify's official top lists from various years. Also in our data set, we will have songs that haven’t been on Spotify’s top list. We will use this data to train our model and test it. Before training our model, we are planning to shuffle the songs to increase accuracy.
We’ll have the attributes of songs such as energy, beats per minute, danceability, duration, release year, dB(loudness), valence, and acousticness.
We plan to use:
Spotipy: For Spotify data collecting
organizeyourmusic: To get attributes of songs
Decision Tree: This Will be used to creating the decision tree.
Random Forest Classifier: Will be used as meta estimator.
Related Papers:
Writers: @mehmetalixk, @gamzedeniz, @cemocancu
