基于鲸鱼优化算法改进长短期记忆神经网络的资源推荐  

Resource Recommendation Based on Long Short Term Memory Neural Network Improved by Using Whale Optimization Algorithm

作　　者：仇焕青 陈曙光[2] 龚芝 张福泉 QIU Huanqing;CHEN Shuguang;GONG Zhi;ZHANG Fuquan(School of Computer Science and Engineering,Hunan University of Information Technology,Changsha 410151,Hunan,China;School of Physics and Electronics,Hunan University,Changsha 410082,Hunan,China;College of Computer and Control Engineering,Minjiang University,Fuzhou 350108,Fujian,China)

机构地区：[1]湖南信息学院计算机科学与工程学院,湖南长沙410151 [2]湖南大学物理与微电子科学学院,湖南长沙410082 [3]闽江学院计算机与控制工程学院,福建福州350108

出　　处：《济南大学学报（自然科学版）》2023年第3期309-315,共7页Journal of University of Jinan(Science and Technology)

基　　金：国家自然科学基金项目(61871204);湖南省教育厅科学研究项目优秀青年项目(22B1028,19B397);福建省科技计划项目引导性项目(2018H0028);湖南信息学院2022年度校级科研项目(XXY022QN04)。

摘　　要：为了改善资源推荐算法的性能,提出基于鲸鱼优化算法(WOA)改进长短期记忆神经网络(LSTM)的资源推荐算法;首先提取资源和用户特征,构建特征差异值加权函数;然后,以资源-用户特征作为输入,建立基于LSTM的资源推荐算法,通过输入门、遗忘门、输出门及记忆节点对历史资源推荐数据按权重进行遗忘与筛选,有选择性地挑选部分数据进行循环迭代训练;考虑到LSTM的门操作需要设置的参数较多,引入WOA进行参数智能优化求解,提出WOA-LSTM算法,以提高LSTM的参数优化的精度及效率。结果表明,通过合理设置WOA参数,可以有效改善LSTM的资源推荐性能,与常用资源推荐算法相比,所提出的WOA-LSTM算法具有更高的推荐精度及稳定性。To improve performances of resource recommendation algorithm,a resource recommendation algorithm based on long short term memory neural network(LSTM)improved by using whale optimization algorithm(WOA)was proposed.Firstly,resources and user features were extracted,and a feature difference value weighting function was constructed.Then,a resource recommendation algorithm based on LSTM was established,resource user features were input,and historical resource recommendation data were forgotten and filtered according to weights through the input gate,forgetting gate,output gate,and memory node.Some data were selectively selected for cyclic iterative training.Considering that there were many parameters to be set for the gate operation of LSTM,WOA was introduced to solve the parameter intelligent optimization and WOA-LSTM algorithm was proposed,so as to improve accuracy and efficiency of parameter optimization of LSTM.The results show that by reasonably setting WOA parameters,the resource recommendation performances of LSTM were effectively improved.Compared with the common resource recommendation algorithms,the proposed WOA-LSTM algorithm has higher recommendation accuracy and stability.

关 键 词：资源推荐 长短期记忆神经网络 鲸鱼优化算法 特征差异值 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]