/Song_Recommender_Project

Primary LanguageJupyter Notebook

Song_Recommender_Project

Team members: Magdalena Blaski-Makara, Muhammad Anani

Project type: Web-scraping / API / K-Means clustering

Aim: Create a user interface that returns song recommendations from databases created by scraping the Billboard Hot 100, and by using the Spotify API

Reasoning:

Work on a topic that we like :) and use API and webscrapping techniques to develope a fool-proof and useful tool; the song recommender.

This project consists of four main parts. Each of these parts is in its own notebook:

  1. Scrapping artist and song names from the "Billboard Hot 100": This then is used in a song recomm
  2. Using Spotify API in order to build a database of songs and their musical features from playlists
  3. K-Means Clustering of song features
  4. Building a Song Recommender that takes user input and then if:
    1. the song is in the top 100; returns to the user a recommendendation from the top 100
    2. the song is NOT in the top 100; returns to the user a song recommendation from the same cluster in the database created using spotify. The user can also play the song using the API imported player.

Data:

Data-Set Features Billboard Hot 100 Spotify Playlists
Columns 2 18
Rows (songs) 100 10270
Source Billboard Hot 100 Spotify API (GB)

Frameowork

Using python:

  1. Scrapping from Billboard Hot 100: this section defines a function that scrapes song and corresponding artist names and concatenates them into a .csv
  2. Building a database of songs with Spotify API: sampling a wide variety of playlists and then adding them all into a .csv file that contains features of songs
  3. Clustering songs from spotify playlists: clustering using K-Means. Choosing the most appropriate number of clusters based on elbow and silhouette methods
  4. Building the song recommender: Finally, build the functions that allow the user to interact with the databases created in order to provide appropriate and fun recommendations

All notebooks all labeled with the corresponding numbers from 1-4 as the bullet points above