摘要
传统网络信息资源个性化推荐方法无法存储长期信息,导致推荐精度低,召回率高。因此,研究基于广义回归神经网络的网络信息资源个性化推荐方法。首先,获取初始兴趣偏好特征数据,分配相应权重进行归一化处理;其次,确定训练样本的收敛范围,调整权值得到不同层神经元之间的连接权值和阈值,并输出匹配结果;最后,运用过滤推荐算法计算环境网络信息资源偏好和用户网络关系,得到训练样本相似度,生成近似数据集,根据偏好完成个性化推荐。实验结果表明,该方法的召回率最低,推荐准确程度高。
Traditional personalized recommendation methods for network information resources cannot store long-term information,resulting in low recommendation accuracy and high recall rate.Therefore,a personalized recommendation method for network information resources based on generalized regression neural networks is studied.Firstly,obtain initial interest preference feature data,assign corresponding weights for normalization processing.Secondly,determine the convergence range of training samples,adjust the weights to obtain the connection weights and thresholds between different layers of neurons,and output matching results.Finally,a filtering recommendation algorithm is used to calculate environmental network information resource preferences and user network relationships,obtain training sample similarity,generate approximate datasets,and complete personalized recommendations based on preferences.Experimental results show that this method has the lowest recall rate and high recommendation accuracy.
作者
吴赟婷
WU Yunting(Jiangxi College of Construction,Nanchang Jiangxi 330200,China)
出处
《信息与电脑》
2023年第5期38-40,共3页
Information & Computer
关键词
广义回归神经网络
信息资源
个性化
推荐方法
generalized regression neural network
information resources
individuation
recommended method