Headquartered in the United States, the client is an American multinational conglomerate holding company and the world’s largest telecommunications company. They provide a range of networking solution and local, regional, long distance, and international telecommunications services to businesses, government entities, and consumers.
The client needs a recommendation system that has a competitive advantage against Netflix on TV and other channels. With more than 20 million subscribers from countries across the globe, our client wants to enhance the user experience by helping them to discover entertainment contents that fit their preferences faster.
FPT creates a recommendation engine that consists of creative logic and algorithms to aggregate, process and analyse the user’s taste. FPT solves the equation by implementing Cloud recommendation to leverage the big data technology and the strength of cloud computing such as AWS, Hadoop, and Spark.
We focus on three popular approaches such as Content-based filtering (recommendation on the basis of a user’s behaviour) and Collaborative filtering (recommendation based on a model of prior user behaviour) to calculate recommendation model; and Clustering (grouping of users’ tastes based on their characteristics and similarities).
Moreover, pre-processing data is also important step that we get involved in. We ingested users’ events from many system devices such as STB, Dotcom, mobile and meta-data of movie.
Processing with big data and testing recommendation system are a big challenge to us because of lacking experience from beginning.
The solution from FPT Software has successfully achieved all goals from the client.
- We construct tailor-made recommendation carousels with relevant contents thanks to the valuable insights and statistics provided from the recommendation engine.
- Personalized interactions show the user that he/she is valued as an individual, thus engenders his/her loyalty.
- Users become more engaged in the service when are able to delve more deeply in it without having to perform search after search.
- Hadoop Mapreduce
- Hadoop Spark