【摘要】：基于深度學(xué)習(xí)的特征抽取是目前數(shù)據(jù)降維問題的研究熱點(diǎn),堆疊自編碼器作為一種較為常用的模型,無法對(duì)混有噪聲及較稀疏的數(shù)據(jù)進(jìn)行良好的特征表達(dá)。面向微博情感分析,通過在堆疊降噪自編碼器的各隱藏層中加入稀疏因子,來解決樣本數(shù)據(jù)所含噪聲和稀疏性對(duì)特征抽取的影響。使用COAE評(píng)測(cè)數(shù)據(jù)集進(jìn)行的情感分析實(shí)驗(yàn)表明所提模型分類的準(zhǔn)確率和召回率都有所提高。
[Abstract]:Feature extraction based on depth learning is a hot topic in the field of data dimensionality reduction. As a common model, stacked self-encoder can not express the features of noisy and sparse data well. For Weibo emotional analysis, the influence of noise and sparseness of sample data on feature extraction is solved by adding sparse factor into each hidden layer of stack de-noising self-encoder. The experiment of emotion analysis using COAE data set shows that the accuracy and recall rate of the proposed model classification are improved.
【作者單位】： 北京工業(yè)大學(xué)信息學(xué)部;
【分類號(hào)】：TP391.1
，

本文編號(hào)：2160790