Team Merlion from IHPC’s Computing Science Department comprising Duan Rubing, Tan Yong Kiam, Hu Nan, Wu Zhen Zhou, Liu Yong, Yang Xulei, Yang Feng, and Rick Goh, clinched 1st prize at the Rakuten-Viki Global TV Recommender Challenge 2015.
The Rakuten-Viki Global TV Recommender Challenge was organized by Dextra, a company under the IDA’s Data Innovation Challenge Scheme. Rakuten owns the largest e-commerce website in Japan and among the world’s largest by sales, and Viki is a Singapore-based video streaming website acquired by Rakuten for US$200 Million. As a key event to celebrate the launch of the Rakuten Institute of Technology hub in Singapore, Rakuten-Viki Global TV Challenge was one of the most anticipated and compelling data challenges hosted in Singapore thus far.
The goal was to build a personalized recommender system for Viki fans worldwide, following a set of user and business considerations, and based on more than 7 million lines of rich, anonymized historical data about users, videos and watching behavior.
Team Merlion from the Computing Science Department at the Institute of High Performance Computing (IHPC), A*STAR, won the 1st Prize out of 567 submissions from 132 participants. The team successfully overcame many challenges posed by the dataset through an innovative 3-step approach. The winning approach utilised different techniques such as classification, filtering and ranking to recommend the most relevant top 3 videos to a user based on the insights derived from the data. Such insights unveiled wide range of viewers’ preferences such as video freshness, genres and geo-information. The proposed recommender system is truly personalised for each individual user.
The team looks forward to collaborate with Rakuten to further refine the technology, and to deploy enhanced personalized recommender technologies for Rakuten.
Congratulations to the team!