Recommender System - Based upon some similarity we will get recommended. like youtube videos, e-commerce recommendation There are three types of recommender system
1. Content Based - Based upon the content we get recommended like similar songs recommended on GAANA.com while listening songs
2. Collaborative filtering - Based upon similar personalities we get recommended. eg two person rated any movie quite similarly. so our algo will treat these two person similarly. And one user will get recommended based upon anothers watchlist.
3. Hybrid based - Both technique is used jointly as per their convenience.
Feature Selection Budget - nobody recommends movie based upon budget Genres - we recommend movies based upon genre link scifi, thriller, comedy Homepage - doesnt required Id - will require for web development keyword - tags..required original_lan - highly imbanced towards english so will drop it original_title - we will use title, original may be in regional language also overview - content depends upon overview, so highly required popularity - numeric value so dont use production_companies - recommendation doesnt based upon production_companies, so dropping production_countries - doesnt required release_date - generation wise, it required but since numeric so dropping it revenue - indirect indicator, but wont recommend based upon runtime- doesnt required spoken_lang - doesnt required tagling - have kept overview so omitting title - will keep it vote_average - numeric will remove it vote_count - again numeric wont use here movie_id - will use id so no movie_id cast - based upon cast we recommend like srk crew - we recommend particular director movie so keep it