Friends’ movie recommendations are welcomed by a lot of film buffs, but sometimes you might want to catch a movie that fits your taste better, based on particular criteria so that you get something that you will almost certainly enjoy. Enter the movie recommendation engine.
Writing in the International Journal of Business Intelligence and Data Mining, researchers from Nigeria have turned to a statistical tool known as Pearson’s correlation coefficient to help them build a new type of movie recommendation engine. Bolanle Adefowoke Ojokoh of the Department of Computer Science at the Federal University of Technology in Akure, Nigeria, and colleagues explain that their approach brings artificial intelligence to personal recommendations. The coefficient allowing collaborative filtering of data based not only on numerical analysis of the data but also the determination of linear relationships among users.
The team tested their approach on datasets assimilate from hundreds of local video shops and information extracted from the Internet Movie Database (IMDb) and ratings by those who have already seen the hundreds of movies analysed. They also added a parental control function to make it child friendly. When they had volunteers test the recommendations the system generated they found that almost 96 percent of users found the recommendations agreeable.
“The system allows new users to be given more personalised recommendations. It also allows users with similar rating patterns to influence the prediction of items,” the team writes. “Our approach offers a more efficient way of managing the cold-start problem in movie recommendation,” they conclude.
Ojokoh, B.A., Aboluje, O.O. and Igbe, T. (2020) ‘A collaborative content-based movie recommender system’, Int. J. Business Intelligence and Data Mining, Vol. 17, No. 3, pp.298–320.