A recommend system for modelling large-scale advertising

In a large-scale network environment, the Internet has produced an amount of digital information such as social media, e-commercial comments, online shopping records and advertising click logs. So, recommend systems not only furnish users with satisfied requirements but also bring about healthy profit to companies. To cope with massive data, we proposed an effective recommend system (ERS) for modelling abundant history data. Firstly, we advise a new data split strategy to make the recommend system fit for large scale advertising. Then, a distributed-recommendation algorithm is designed. Finally, we apply ERS and distributed computing and storage tools (such as Spark, Hadoop, etc.) to model history data aiming at improving the system efficiency. In this paper, we also compare and analyse diverse performances of ERS including its efficiency and its accuracy.