CAJ | 학술논문

根据用户搜索历史，将用户关注的信息按标题分类，通过自编码神经网络提取特征值。设定学习样本标题最多为25个汉字，编码方式采用汉字机内码(GBK码)。使用 MATLAB工具进行深度学习，将样本在原空间的特征表示变换到一个新的特征空间。
근거용호수색역사，장용호관주적신식안표제분류，통과자편마신경망락제취특정치。설정학습양본표제최다위25개한자，편마방식채용한자궤내마(GBK마)。사용 MATLAB공구진행심도학습，장양본재원공간적특정표시변환도일개신적특정공간。
According to user search history,the user information of interest was classified by Title,from which the fea-ture value was extracted by the auto-encoder neural network.The method set the learning sample heading up to 25 Chinese characters,adopted Chinese characters machine code (GBK code)to realize coding mode,used the MATLAB tool for deep learning,and transformed the feature in the original space representation into a new feature space.