Song Recommender Project | Unsupervised Machine Learning

Project Overview

The aim of this project was to build an automated song recommender and is structured in multiple steps:

First, the Top Currently Trending Billboard 100 Songs were webscraped from the respective website and saved to a csv file [BB_Hot100_Webscraping.ipynb].
A 43000 entry dataset [curated_audio_features_dataset_curated.csv] of songs and their respective audio features (such as tempo, loudness, danceability) was the basis for applying unsupervised machine learning to cluster them based on similar song audiofeatures. The best results were achieved through a grouping in the following 7 clusters: 0: "Dancy", 1: "Rocky", 2: "Classy", 3: "Folky", 4: "Electry", 5: "Poppy", 6: "Rappy" and stored in a new CSV file [new_af_clusters_7.csv].
The song recommender function works by asking the user what mood they are in today and offers the predifined clusters options as well as an additional "trendy" category based on the previously scraped top one 100 songs. It the randomly selects and plays one song from the respective cluster/ group. [Song_recommender.ipynb]

Name		Name	Last commit message	Last commit date
Latest commit History 8 Commits
.ipynb_checkpoints		.ipynb_checkpoints
BB_Hot100.csv		BB_Hot100.csv
BB_Hot100_Webscraping.ipynb		BB_Hot100_Webscraping.ipynb
New_Clustering_Audio_Features_Curated.ipynb		New_Clustering_Audio_Features_Curated.ipynb
README.md		README.md
Song_recommender.ipynb		Song_recommender.ipynb
curated_audio_features_dataset_curated.csv		curated_audio_features_dataset_curated.csv
new_af_clusters_7.csv		new_af_clusters_7.csv