Skip to content

README.md

Latest commit

 

History

History
31 lines (24 loc) · 1.63 KB

README.md

File metadata and controls

31 lines (24 loc) · 1.63 KB

iTunes-SQL-Machine-Learning

Using MySQL to perform ML prediction of song rating on my iTunes Music Library

The purpose of this project is to produce a trained model that can predict a song rating for the tracks in my iTunes music library.

The main steps are:

  1. Upload iTunes Music Library into an SQL database.
  • This process is outlined in the "iTunesSQL" document.
  1. Use a set of already rated tracks to train and test 6 different models
  2. Use the best model on my full iTunes library to rate my unrated songs.

I wanted to explore how well the models would predict rating based purely on a ratio of (Play_Count÷Skip_Count), so I generated a test dataset with ratings based on this ration plus some noise.

To generate the test dataset:

  1. I uploaded a sample of about 300 songs.
  2. I filled in missing Play_Count and Skip_Count with random integers.
  3. To unrated songs, I assigned a Rating value based on 3 different ranges of the ratio Play_Count/Skip_Count.

Assignment of ratings based mainly on ratio of Play_Count / Skip_Count: To introduce some variability, I assigned the Ratings

  1. For Ratio <=0.5, Rating= [random sequence of 0 and 40]
  2. For 0.5 < Ratio < 2.0, Rating= [random sequence of 40 and 80]
  3. For Ratio >=2.0 , Rating= [random sequence of 80 and 120]

To train the models:

  1. I randomly split the test dataset into 80% training set and 20% validation set.
  2. I provided the models with Play_Count, Ratio, Total_Time (song duration), and Rating.
  3. Set up a 10-fold cross validation test harness.
  4. Studied the confusion matrix and classification report.

The trained model then be used on my full iTunes Library!