Recommendation system built on data scraped from Rotten Tomatoes
Executive Summary
Over the last decade, most of the world's biggest entertainment and telecom have moved enormous attention to the new battlefield of streaming entertainment. From Statista's report of Frequency of streaming movies in the U.S. 2019-2020 (Watson, 2020), there are a total of 25 percent of adults who aged between 18 and 29 years old and a total of 9 percent of adults who aged above the age of 65 years old claimed that they watched movies every day. In response, there are more than 750 movies released in the United States and Canada in 2018 and 2019 (Watson, 2020). This growth of online movie consumption has brought a tremendous amount of information and choices to consumers and requires a better movie recommendation system to help consumers find the right movies they want. In general, content-based and collaborative Filtering recommendation system are the two most popular approaches.
Content-based and collaborative filtering recommendation both rely on the item-user interaction. Beside using users' interactions and feedbacks, the content-based recommendation also requires rich information of features of the movie to find new movies that are similar to what the user has watched and liked in the past. On the other hand, collaborative filtering recommendation requires a good amount of users' information that helps find similar users and create a recommendation list based on what a similar user enjoyed.
However, both recommendation systems have to face three problems that are: cannot recommend fresh items, cold start problem, and lack of side features for the query. In this project, we present a new approach to designing a new content-based recommender system to overcome these problems.