Song Popularity Predictor


Prediction Model Types

Logistic Regression

Accuracy: 65.2%

K Nearest Neighbors

Accuracy: 61.2%


Accuracy: 68.7%


"Music is an undeniable part of our lives"

"Music is an undeniable part of our lives - we listen to music during commute, work, or even to get over a heartbreak. The best ones end up in the Billboard 200 chart, but here's a central question:
Is there a method to this madness that is being a chart-topper?

The objective of this project is to build a model that can predict the potential of a song to be a billboard hit based on its features.
This project involves the following:
1. API calling and web scraping - Python3
2. Data cleaning - Python
3. Data reporting - Jupyter Notebook
4. Machine learning - Python


Our Project Journey

  • 2nd April 2019

    Project commences!

    The “Project Billboard” team was formed. The task was to deliver a model predicting if a song would be a Billboard hit.

  • 18th April 2019

    Exploring possibilities

    We had our first meeting at starbucks where we worked out the scope of the project and the resources to use. Spotify’s API was chosen as our primary source of data for song features.

  • 29th April 2019

    Song extraction & analysis

    The code for collecting song features from Spotify was completed and we had our first dataset. We proceeded to try simple algorithms for hit/miss prediction, as well as scraping billboard hot 100 charts for unique hits.

  • 6th May 2019

    Model training & refining

    Machine learning models proved decent, with accuracies of 60-70%. The next step was to explore how we can refine the model and prepare a user-friendly interface.

  • 11th May 2019

    Project deployed!

    Our models have been deployed on a simple website with a user-friendly interface!

Project Team


Tan Jin

Tan Jin

The Cool Kid

Arthur Chionh

Arthur Chionh

The Funny Kid

Toh Yue Feng

Toh Yue Feng

The Crazy Kid

Have an idea for the project? Reach out to any of us!