Research On Recommendation Methods Of Scientific And Technological Papers Based On T-SNE And Fuzzy Clustering

The rapid development of the Internet in recent years and its important carrier of the low cost of storage and convenient query capability makes it become the scientific papers.At present,the platform of science and technology paper retrieval is basically based on text retrieval technology.The desire of scientific research users to obtain the scientific papers they need quickly and accurately on a search platform that contains numerous scientific and technological papers is still difficult to achieve.Therefore,many researchers will focus on the personalized science and technology paper recommendation field.Most of the scientific papers on the Internet exist in the form of text.The most convincing factor in judging whether the two texts are similar is nothing more than the content of the text.In view of the problem of neglecting this important factor in the traditional science and technology paper recommendation field and the waste of time and space caused by the comparison of scientific and technological papers in the database,this paper begins with the content of the scientific and technological paper.A method of recommending scientific papers based on t-SNE and fuzzy clustering is proposed.Using the advantage of t-distributed neighborhood embedding(t-SNE)algorithm in dealing with high-dimensional data,reducing dimensionality of modeled scientific paper collection matrix,and uses fuzzy clustering algorithm to cluster t-SNE processed data to realize the personalized recommendation of scientific and technological papers based on t-SNE and fuzzy clustering.First,the space vector model is used to model scientific papers.Secondly,in order to solve the problem of the excessive dimension of the model brought by the vector space model,a new algorithm called t-SNE algorithm is proposed to reduce the dimension of the scientific and technological paper model.Then,considering the intersection between science and technology thesis subject,a fuzzy C-means(FCM)algorithm is proposed to cluster scientific papers,which avoids the hard clustering caused by hard partition problem to ensure students of cross characteristics,but also reduce the judgment of similarity of unnecessary comparison between scientific papers.At the same time,aiming at the uncertainty of the clustering result caused by the FCM,a simple algorithm is proposed to determine the clustering number automatically.Finally,the content-based recommendation method is used to make personalized recommendation for scientific research users.The experiment show that the proposed method is superior to the traditional scientific papers recommendation algorithm in time and space complexity.At the same time,the accuracy of the recommendation is also improved compared with the traditional science and technology paper recommendation.