Recommender System with Machine Learning and Artificial Intelligence

10Content (Item)-Based Recommendation System

R. Balamurali

Dept. of Computer Science & Engineering, IcfaiTech, IFHE, Hyderabad, India

Abstract

The main idea behind Content-Based Recommendation System is to recommend an item to a customer similar to the previous items that are rated high by the same customer. Initially, we create an item profile based on the ratings provided by the user for the items he purchased, videos/movies watched, books (textual contents) read, music/songs heard etc.. The user profile captures information like frequently purchased products/brands, type of movies watched (with a specific actor/director/genre etc...), type of music listened (melody/beat etc...) and type of books read (comic/positive thinking etc...) by the particular user. The item profile will be represented as a Boolean vector based on the features of interest for a given user. The item profile will be created for all the items rated by a specific user. From, the item profile we infer the user profile. A simple way to create user profile is to take the average rating given by the user for a specific item. There are also several other ways of creating user profiles which we will discuss in this chapter in detail. Basically, the user profile captures the details of items which are likeable by a particular user. Once the user profile is created, given a new item that is not rated by the user which may be a product/movie/song/book etc.…, it can be represented as a Boolean vector and this Boolean ...