MusicRecommendationSystem

We built a recommendation program for “music”. The user enters in two songs and the main class ‘Music_Recommender’ recommends a list of songs relative to the entered two songs.

Class ‘Music_Recommender’ calls other three classes that handles the following three filtering models :

• Collaborative Filtering – KNN Classifier
• Content Based Filtering – Lyric based Cosine Similarity
• Content Based Filtering – Music Features based Sigmoid calculation

Program is also capable of retrieving new dataset from server by ‘Crawl_melon(self):’ function.

Music_Recommender

' A Top-Level Class Music_Recommender is an all-in-one model music recommender Class. You can call different Models to get music suggested, used models: '

• Collaborative Filtering – KNN Classifier
• Content Based Filtering – Lyric based Cosine Similarity
• Content Based Filtering – Music Features based Sigmoid calculation

init ( self, df_KNN_count, df_KNN_meta, df_Lyric, df_Feature):

Initializing Function. Enter datasets provided for each filtering method. Preprocesses The Dataframe inside this function.

KNN_Recommend(self, song1, song2):

Function that calls Collaborative Filtering Model. Returns a list containing the recommended songs

CB_Lyric_Recommend(self, song1, song2):

Function that calls Content Based Filtering – Lyric based Cosine Similarity Returns a list containing the recommended songs

CB_Feature_Recommend(self, song1, song2):

Function that calls Content Based Filtering – Content Based Filtering – Music Features based Sigmoid calculation Returns a list containing the recommended songs

Crawl_melon(self):

Function that execute crawling on the website : “www.melon.com” Saves the dataframe to "melon_top100.csv"

Collaborate_Recommender:

Collaborate_Recommender is a class that handles KNN and Content based Filtering.

init (self, metric, algorithm, k, data, decode_id_song):

Initializing Function. Enter datasets provided for each filtering method.

make_recommendation(self, new_song1, new_song2, n_recommendations):

Creates recommendation for two songs. Returns a list containing the recommended songs

_map_indeces_to_song_title(self, recommendation_ids):

Internal function that gets reverse mapper

_fuzzy_matching(self, song):

Initiates Fuzzy Matching

CB_Lyric_Recommender:

CB_Lyric_Recommender is a class that handles Content based filtering that uses lyric data.

init (self, metric, algorithm, k, data, decode_id_song):

Initializing Function. Enter datasets provided for each filtering method.

_print_message(self, song, recom_song):

Prints out the result of the recommendation

recommend(self, recommendation):

Function that gets recommendation from 1 song Returns a list containing the recommended songs

recommend2(self, recommendation):

Function that gets recommendation from 2 song Returns a list containing the recommended songs

Reference : Dataset : https://www.kaggle.com/datasets/maharshipandya/-spotify-tracks-dataset http://millionsongdataset.com/

Data Crawling : https://dada-devdiary.tistory.com/34?category=994161

Content-based-filtering : https://towardsdatascience.com/the-abc-of-building-a-music-recommender-system-part-i-230e99da9cad

Collaborative-filtering : https://github.com/ugis22/music_recommender