Research On Recommender System For Big Dataset

Cloud computing, Internet of things, intelligent terminal device and other newtechnology have been studied a lot over the last decade. Many achievements have beenmade both in the industry and academia, which leads us to the age of big data. In thisbackground, the number of users and their recommended products is increasingsignificantly. Traditional collaborative filtering recommendation method can not workwell with big user-item rating dataset. It’s important to propose a new architecture forrecommender system and improve the traditional collaborative filtering recommendationmethods.First of all, based on the research of current approaches for recmmender system andbig dataset, we find that It is necessary to develop a recommender system which is reliablescalable and also can support a large number of users and items. After collecting andorganizing various related information, this paper established the architecture of bigdataset recommender system which contains three layer: user interaction layer,recommendation engine and cluster layer and then make all the detail designs.Secondly, traditional collaborative filtering recommendation methods suffers fromdataset sparse, cold start and efficiency problems and recommend accuracy decreases withthe increase of the amount of data. Therefore, we improved the traditional collaborativefiltering recommendation method by increasing the same rating between two user whencalculate their similarity and running it on the cluster. Because of above actions, thecollaborative filtering recommendation method obtain a better accuracy.Finally, we made a experiment in this paper and draw a conclusion on the results thatthe method we proposed is better than traditional collaborative filtering recommendationmethods in accuracy and efficiency.