| Over the past decade,the development of the Internet has brought the development of network and social civilization to the new era of information.Growth data provide users more choice,but how to choose their information demand has become increasingly difficult.In the face of vast amounts of data,both for the user and supplier for how to get interested in things has become a great challenge,user is difficult to get their needs in data of Marine information,supplier is difficult to analysis in the large amount of information garbage out the user’s real needs to elevate their interest.The amount of information from the ocean is extracted and refined to accurate and precise information recommendation system plays an important role.Collaborative filtering can directly or indirectly tap the user’s explicit or implicit demand information and generate recommendations.In the practical application process,but often due to data sparsity,cold start problems can not get a better recommendation effect.Therefore,this paper to the neighbors of the similarity measure and target user choice as a starting point,it was put forward that the collaborative filtering algorithm based on cloud model of Euclidean space similarity of collaborative filtering algorithm and user attributes weighted active neighbor.The Contents are as follows:Firstly,Reviewed the basic principle of collaborative filtering algorithm and cloud model of collaborative filtering algorithm,this paper introduces the basic principle and process of cloud model collaborative filtering recommendation algorithm based on cosine similarity measure.In addition,the research and Analysis on how to generate the nearest neighbor set of the target user in the collaborative filtering algorithm based on k NN nearest neighbor.Secondly,In the collaborative filtering recommendation algorithm based on cloud model,cosine similarity measuring method encounters the problemsoffeatures missing and lack of discrimination and not take into consideration the length of vectors.In view of this,we propose a new method which is using Euclidean distance to measure similarity,bythree digital features of cloud model which are cloud expectation,entropy and hyper entropyviewing as a three-dimensional space coordinates.We use Euclidean distance to measure the similarity of cloud feature vectors to achieve the qualitative representation of data.Experimental results show that the new similarity measuring method can not only improve the differences of cloud feature vector but also provide better recommend quality of the collaborative filtering recommendation based on cloud model.Thirdly,Aiming at the problem that the nearest neighbor’s collaborative filtering technology based on k NN is extreme dependence on the rating similarity in the choice of nearest neighbors,a user-attribute-weighted active k-nearest neighbor’s collaborative filtering algorithm was proposed.Firstly,the user’s feature attributes were introduced and fused minimum weight similarity.k NN nearest neighbors set of the target users were generated according to the final similarity.Active user’s subpopulations of target items were generated from users who had feedback from the target items.The active user’s subpopulations of k NN nearest neighbors were selected as the active nearest neighbor sets of target users.Finally,score predicts were produced.Experimental results on the public data sets show that the proposed algorithm can effectively improve the recommendation accuracy and has better stability. |
