In this project, I utilized the "Spotify Tracks Dataset" available on public domain at Spotify Tracks Dataset. 😊✨
I used standard machine learning techniques such as data exploration and visualization, followed by statistical methods such as doing a PCA and a correlation heatmap analysis. Afterwards, I created a recommendation system that fetches the most similar track to the user's chosen track. Additionally, I compared classification methods based on audio features provided by the dataset such as Random Forest, Gradient boosting, FNN and MLP. I concluded the analysis by analysing one property - "popularity", in order to find what makes a track popular and what influences this. Lastly, I did a genre-based analysis by finding the most energetic, most danceable genres and genres with the highest valence. 🎛️🎶
For this project, I've used the following libraries: Pandas, Sklearn, Numpy, Tensorflow, Keras, Plotly, Seaborn, Matplotlib.
I also created a report on the findings of an in depth analysis to accompany the code. The code is available on a Google Colab notebook and a Python script.
Important
To view the full project, I'd personally suggest you to please view the Jupyter notebook titled spotify.ipynb file for the code with commentary and visualizations for the best experience.