• Using Spotify API

• Collect ~25000 songs in Spotify’s database

• ‘Spotifyr’ library in R

• 5 features which describe a particular song

  1) Acousticness
  2) Danceability
  3) Liveliness
  4) Loudness
  5) Speechiness

• Every feature has a scale which varies from 0 to 1 except loudness.
• Normalizing Loudness feature as it has a scale from -60db to 0db
• Remove rows that have Nans and will use logical indexing to remove rows with a value of 0.

• To understand the mood of a listener to recommend tracks based on their current mood
• Results in enhancing the listener’s user experience
• Example: imagine if a person gets suggestion to listen to romantic music when he is angry, it may lead to a bad user experience

Interesting part:

• We can infer the mood of a listener based on his/her recent musical history

• Classify songs by mood by using a clustering algorithm
• Unsupervised learning algorithm
• K-Means Clustering Algorithm
• Elbow Method to decide number of clusters in our data

-Qualitative Results:
1.Predict the mood of a person provided the song they are listening
2.Visualize and plot clusters using ggplots and other graph packages.

-Quantitative Results:
1. Confusion matrix
2. Least error for clusters
3. F1 Scores - harmonic mean of the precision and recall