Research On Personalized Recommendation System Of Agricultural News Based On Emotion

With the rapid development and popularization of the Internet and smart phones,reading online news articles reporting agricultural messages has become the most important way for people to obtain agricultural information required.However,a serious problem called“information overload,caused by the Internet restricts the public from accessing to agricultural information timely and effectively.The personalized recommendation system is one of important ways to solve information overload significantly.Although existing achievements perform well,the shortages still remain.For instance,the problems of cold start and low recommended quality resulted in by sparse data and inaccurate sentiment labeling and classification of news articles.Therefore,this thesis focuses on the problem of recommending personalized agricultural news articles with high quality based on sentiment.The research contributions are as follows.(1)Sentiment classification of agricultural news articles based on Joint Sentiment-Topic Model(JST in short).First,the feasibility of JST applied to agricultural news data is verified.Then,the sentiment dictionary including positive and negative keywords is constructed by combining the uniqueness of agricultural field,which is used for improving the precision of sentiment classification.Finally,an updated JST blending the prior knowledge of agriculture is proposed to automatically classify the sentiment of agricultural news articles with high preciseness.(2)Algorithms for recommending personalized agricultural news articles.The observation reports that similarity between users who like a same project with unpopularity or professional is higher than that between users who like project with high heat.Hence,a user-interest model is firstly built based on attributes of users and items' types,and then a novel similarity algorithm,namely L-HUMCF,is proposed by importing hot factor.The algorithm can significantly improve the accuracy of selecting the neighbor set of a user.In reality,users rate items subjectively.Thus,to measure the effectiveness of users' ratings and to reduce the selection bias of users' neighbor sets,a novel personalized recommendation algorithm importing heat factor and entropy,namely EL-HUMCF,is proposed to further improve the recommendation quality of agricultural news articles.(3)Design and implementation of the personalized agricultural news recommendation system based on sentiment.A sentiment-based personalized agricultural news recommendation system is developed by utilizing Java language and MySQL database.The developed system can recommend personalized agricultural news articles with high quality to users and provide them better reading experience.Meanwhile,by recording the sentiment classifications of the recommended agricultural news articles,the developed system may infer side information to furtherly improve the recommendation quality.