兴趣深度融合下图书多特征过滤推荐目标仿真  

Simulation of Book Multi Feature Filtering Recommendation Targets under Deep Fusion of Interest

作　　者：于敏 孙小飞[2] 江婕[3] YU Min;SUN Xiao-fei;JIANG Jie(School of Mathematics and Computer Science,Yuzhang Normal University,Nanchang Jiangxi 330103,China;PLA Naval Submarine Academy,Qingdao Shandong 266199,China;Jiangxi Science and Technology Normal University,Nanchang Jiangxi 330038,China)

机构地区：[1]豫章师范学院数学与计算机学院,江西南昌330103 [2]海军潜艇学院,山东青岛266199 [3]江西科技师范大学,江西南昌330038

出　　处：《计算机仿真》2024年第9期433-436,525,共5页Computer Simulation

基　　金：2021江西省教育科学十四五规划一般课题《运用知识建构理论培养师范生批判思维的实证研究》(21YB281);2020江西省教育科学十三五规划重点课题《基于STEM理念的大学生创新创业教育实证研究》(20ZD049)。

摘　　要：由于用户的评分行为通常是非常稀疏的,即用户与图书之间的交互数据不够丰富,使得难以准确挖掘用户的兴趣和偏好,导致图书的推荐效果并不理想。为此,提出一种考虑兴趣深度融合的图书多特征过滤推荐仿真方法。采用渐进遗忘函数,获取兴趣特征权重,根据获取的借阅次序和时间对特征词权重的影响,获得图书特征词权重大小;定位用户地理位置,将词向量拟作输入数据,把情感取向拟作输出数据,通过卷积神经网络筛选用户网络签到信息,获取用户发布兴趣偏好的概率函数,将兴趣深度融合,计算图书推荐项目的冷门度,根据相似度将用户划分至对应簇中,引入时间跨度因子,采用协同过滤推荐方法,完成图书多特征过滤推荐目标。仿真结果表明,所提方法推荐平均绝对误差值处于0.2左右的水平,用户体验度得分高,很好地提升了图书过滤推荐效率与精准度。Due to the sparse nature of user rating behavior,which means that the interaction data between users and books is not rich enough,it is difficult to accurately explore users' interests and preferences,resulting in unsatisfactory book recommendation results.Therefore,a simulation method for book multi feature filtering recommendation considering deep interest fusion is proposed.Firstly,a progressive forgetting function was employed to calculate the interest feature weights.Then,the influence of the borrowing order and time on the feature word weights was used to obtain the weight of the book feature words.Secondly,the user's geographical location was identified.Moreover,the word vector was used as input data,and the emotional orientation was taken as output data.Furthermore,convolutional neural networks were used to filter users' network check-in information,thus calculating the probability function of the interests and preferences published by users.Next,interests were integrated in depth.Meanwhile,the recommended item which was not in popular demand was calculated.Based on similarity,users were divided into corresponding clusters.Finally,the time-span factor was introduced,and then a collaborative filtering recommendation method was adopted to complete the goal of book multi-feature filtering recommendation.The simulation results show that the average absolute error value of the proposed method is around 0.2.And the user experience score is high.Therefore,this method can effectively improve the efficiency and accuracy of book filtering recommendations.

关 键 词：兴趣深度融合 多特征 过滤推荐 卷积神经网络 权重计算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]