Project Title: Music Classification Model for Musical Fun
Group Members: Valerie: AzureML Qinqin: HTML and Heroku Belinda: Python and Pandas Celine: HTML and JavaScript
PROJECT OVERVIEW
The overarching goal of this project was to develop a model that can accurately classify music based on our dataset and Spotify API and correctly output the music according to categories created.
We were curious to find out how music classification models can do so for users, other than users listening history.
Data Source Kaggle Spotify API
Tools Used • AzureML • HTML • Python • Pandas • Matplotlib • JavaScript • Heroku
We got four sets of data from Kaggle. We cleaned the datasets by dropping null values and merged all four datasets to one.
Once we read in the data in AzureML we elected Multiclass Decision Forest model to train and use it to predict the tracks in to four categories, resulting to a predictive average accuracy score of 0.916949.
We used python and pandas to view our dataset and plot graphs. We chose to use music features to classify our dataset in to four music groups which are workout, party, dinner, and sleep. The cluster visualization show that our dataset was accurately trained.
We created an interactive website whereby the user can input music feature values for the model to predict/ recommend which group the track falls in to, we used HTML and JavaScript. Lastly, we hosted the website in Heroku.