A Personalized Recommendation Model Based On Expert Opinion

Personalized recommendation is a new hot spot. Personalized recommendation is the behavior which is recommending items for targeted consumer by a specific recommendation model based on the information of user buying items, or the historical data of user browsing items. Some existing recommendation models calculated the information which consumers rated items, sorting the result, and then recommended it. In the past years, recommendation system studies have yielded some results, with successful applications of personalized recommendation system in companies such as Amazon, as well as Internet technologies updating. Researchers innovate not only in collaborative filtering and content-based filtering, but also in personal recommendation on the network structure.For the rating of items, consumers are not as professional as experts in this field. Considering the advantage of experts in his field, a personalized recommendation network model based on expert opinion, recommending for the common consumers, are proposed in this paper. The new model is no obvious loss on the accuracy and good diversity on the recommendation result, so it is both diversity and accuracy. The simulation experiment of the new model is on Movielens dataset and Rotten Tomatoes dataset, by the way the two datasets are two completely different data sources and not directly associated with each other. The result shows that, datasets from different sources can be recommended for each other, and for a large dataset, a relatively small dataset can be used to forecast the consumer recommended, significantly reducing the amount of calculation in the recommendation process.