LSTM-GAN:融合GAN和Bi-LSTM的無(wú)監(jiān)督時(shí)間序列異常檢測(cè)
小型微型計(jì)算機(jī)系統(tǒng)
頁(yè)數(shù): 9 2022-11-28
摘要: 多元時(shí)間序列數(shù)據(jù)的異常檢測(cè)旨在發(fā)現(xiàn)對(duì)應(yīng)時(shí)序特征中不符合一般規(guī)律的特異性模式,識(shí)別特定時(shí)間步長(zhǎng)中的異常狀態(tài).針對(duì)多元時(shí)序數(shù)據(jù)時(shí)間依賴性建模難以及數(shù)據(jù)維度不斷增加導(dǎo)致難以有效進(jìn)行異常檢測(cè)等問(wèn)題,本文以自編碼器為基礎(chǔ),融合生成對(duì)抗網(wǎng)絡(luò)(Generative Adversarial Network, GAN)和雙向長(zhǎng)短期記憶神經(jīng)網(wǎng)絡(luò)(Bi-directional Long Short-...