Definition and History of Recommender Systems

By: Nunung Nurul Qomariyah, Ph.D

Recommender Systems (RSs) is an emerging research field that has grown fast and become popular. The increase of interest in this research topic has also been driven by great improvements in internet technology and e-commerce. RSs have many advantages for e-commerce. Three ways in which RSs can enhance an e-commerce system, i.e. by helping buyers with no experience in online shopping, by cross-selling the products and by improving customer loyalty. The peak explosion of research in RSs occurred when Amazon launched their Collaborative Filtering (CF) method at the end of the 1990s, successfully increasing their sales. The successful Amazon became popular and other online businesses started to implement RSs on their website. Amazon has patented its CF method as a United States Patent. Due to the fact that the main goal of an RS is to find the preferred information and eliminate information which is not liked by a user, the RS field can be considered as a subset of information filtering. The process of exploring a user’s preferences from their historical data is followed by processing it using machine learning algorithms to build a ranked list of recommended items, as preferred by the user.

The idea of exploiting computers to recommend the best item for the user has been around since the beginning of computing. The first implementation of the RS concept appeared in 1979, in a system called Grundy, a computer-based librarian that provided suggestions to the user on what books to read. This followed in the early 1990swith the launch of Tapestry, the first commercial RS. Another RS implementation for helping people find their preferred articles was launched in the early 1990s by GroupLens, a research lab at the University of Minnesota, USA. They named the system after the group, GroupLens Recommender System. This system claims to have a similar spirit to that of Tapestry, Ringo, BellCore and Jester. Further development of RSs in the late1990s was the implementation of Amazon Collaborative Filtering, one of the most widely known RS technologies. Since this era, RSs based on Collaborative Filtering has become very popular and has been implemented by many e-commerce and online systems. Many toolboxes for RSs have also been developed. The success story of Amazon also gave rise to the development of many RS algorithms known as hybrid approaches, which combine multiple approaches.

Following the successful era at the end of the 1990s, the industry offered generous funding to implement RSs research. The most popular competition in RSs was held by Netflix, a provider of internet streaming media. They launched the Netflix Prize1in 2006 and give1 million US Dollars to the winner of the competition who provided the best RS movie recommendation. They announced the winning team in 2009. In 2010, YouTube also implemented an RS on its website.

Sources:

Qomariyah, N. N. (2018). Pairwise Preferences Learning for Recommender Systems (Doctoral dissertation, University of York).