三角模糊数的相似度求解在社交推荐系统中的应用研究

北京理工大学珠海学院2020届本科生毕业论文The application of similarity of triangular fuzzy Numbersin social recommendation systemAbstractTraditional social recommendation algorithm can have the problem of insufficient data,inorder to alleviate the traditional social recommendation algorithm under the condition of lackof the original data of the existing calculation method of low performance and socialinformation,and to make society recommend to further improve the accuracy ofrecommendation algorithm,in this paper,the social information in the original algorithm,theoptimization calculation,obtained a new multilevel mixed triangular fuzzy similarity,and thenthe similarity in the information society,as a social recommendation algorithm to get a highersocial recommendation algorithm accuracy.The key of the algorithm in the calculation of social information,social informationcalculation mainly consists of three different levels of similarity,the:the start is the user's scoreas a breakthrough point,using the concept of triangular fuzzy number to the user's ratings havebecome blurred,then calculate the user's triangular fuzzy similarity,and in addition there aretwo kinds of similarity to auxiliary,respectively is the preference and user rating ofuser ratingsJarccad similarity,the final combination as the user rating similarity;The second similarity isto classify the project category,and predict the user's interest in different project categoriesbased on the user's score of the project,so as to calculate the interest similarity between differentusers.The last similarity is to use the unique characteristic attributes in each user to calculatethe characteristic similarity between different users.Finally,the similarity of the three levels iscombined to form the multi-level triangular fuzzy hybrid similarity,and the social informationneeded in the social recommendation algorithm is formed.Used in common use in the current paper MovieLens public data sets,to improve thealgorithm,and finally it is concluded that in the course of the algorithm of parameterdetermination and contrast experiment,the results show that when the dynamic weights of fusedscore of 0.40,the improved hybrid algorithm relative to baseline user similarity matrixdecomposition test set average absolute deviation(MAE)fell by around 4%.In this paper,thealgorithm adopts the strategy of multi-level triangular fuzzy similarity to improve the accuracyof the algorithm for user recommendation,and by integrating more social information,iteffectively alleviates the problem of data sparsity and the influence of the single measurement