A project for CSC 422 Automated Learning & Data Analysis at NC State.

We hoped to be able to predict an albums popularity on the year end Billboard top charts based on various acoustic features. Our models assumed an album was popular if the rank was ≤ 25 or not popular if the rank was > 25.

In order to assess whether or not an album is popular, we utilized different machine learning models:

• Naive-Bayes
• Decision Tree (utilizing Information Gain and Entropy)
• Support Vector Machine
• Deep Neural Networks

• Full Album Data with Acoustic Features (Link to Dataset)

Created using data from:

• Acoustic and meta features of albums and songs on the Billboard 200
• The Billboard Year End Top Albums List

The first dataset was used for the acoustic features and the the Top Albums List was scraped for the album name

We performed a 70/30 training testing spilt and standardized the data

• Link to Training Data
• Link to Testing Data

1. Make sure you have installed virtualenv, or if not then run pip3 install virtualenv
2. Create the python three virtual environment virtualenv venv
3. Start the environment source venv/bin/activate
4. Automatically install all relevant dependencies using the following command pip install -r requirements.txt

Allow dataset_download.sh permission to execute by running

$ chmod +x dataset_download.sh

Download the data byt running

$ ./dataset_download.sh

The training and testing data should be available in data/

In the root folder of the program run this command to start the virtual environment

$ source venv/bin/activate

After the virtual environment has started run this command to start the program

$ python models/decision_tree.py
$ python models/knn_model.py
$ python models/naive_bayes_model.py
$ python models/neural_net.py
$ python models/svm_model.py

Data Science for Hit Song Prediction

Song Popularity Predictor