基于深度神经网络的继续教育信息智能推送技术研究  

Research on intelligent push technology of continuing education information based on deep neural network

作　　者：胡悦 罗小楠 王彬 张伟 HU Yue;LUO Xiaonan;WANG Bin;ZHANG Wei(Teaching Evaluation Center,Air Force Medical University,Xi’an 710032,China)

机构地区：[1]空军军医大学教学考评中心,陕西西安710032

出　　处：《电子设计工程》2021年第14期42-46,共5页Electronic Design Engineering

基　　金：陕西省重点研发计划项目(2018YBXM-SF-17-5)。

摘　　要：在线教育因其系统化、海量化、便捷化以及可以实现精准推送服务而被广泛应用。传统的课程推送模型通常依靠深度学习算法进行模型训练,利用其特征提取学习能力对系统的隐藏特征进行提取。在解决抓取数据过程中出现稀疏问题的同时,也对计算机性能提出了更高的要求,模型训练需要耗费大量的时间。文中针对上述传统课程推送模型算法的不足,将多层神经元模型算法与个性化推荐算法进行结合,并对课程数据进行训练验证。实验结果表明,文中所建立深度神经网络推荐模型的MAE值相较于传统算法有明显的下降,同时训练迭代次数也大幅减小,说明提出的算法模型可以有效缩短数据的训练时间,提高模型的推荐性能。Online education is widely used because of its systematic,quantitative,convenient,and accurate push services.Traditional course push models usually rely on deep learning algorithms for model training,and extract hidden features of the system by using its feature extraction learning.While solving the problem of sparseness in the process of capturing data,it also puts forward higher requirements on computer performance.Model training takes a lot of time.Aiming at the shortcomings of the above-mentioned traditional course push model algorithm,the article combines the multi-layer neuron model algorithm and the personalized recommendation algorithm,and trains and verifies the course data.The experimental results show that the MAE value of the deep neural network recommendation model established in the article is significantly lower than that of the traditional algorithm,and the number of training iterations is also greatly reduced,indicating that the algorithm model proposed can effectively shorten the data training time and improve the model recommended performance.

关 键 词：深度学习 多层神经元网络 推荐算法 继续教育 MAE值 模型训练 

分 类 号：TN99[电子电信—信号与信息处理] TP391[电子电信—信息与通信工程]