• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ÇÐȸÁö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ÇÐȸÁö > µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

Current Result Document : 6 / 11 ÀÌÀü°Ç ÀÌÀü°Ç   ´ÙÀ½°Ç ´ÙÀ½°Ç

ÇѱÛÁ¦¸ñ(Korean Title) µö·¯´×À» ÀÌ¿ëÇÑ ´Ùº¯·® ½Ã°è¿­ µ¥ÀÌÅÍÀÇ °áÃøÄ¡ 󸮿¡ °üÇÑ ¿¬±¸ Á¶»ç
¿µ¹®Á¦¸ñ(English Title) A Survey on Handling Missing Values in Multivariate Time Series Data Using Deep Learning
ÀúÀÚ(Author) ÀÌ¿µÁØ   Young Jun Lee   À±¼ö½Ä   Susik Yoon   ÀÌÀç±æ   Jae-Gil Lee  
¿ø¹®¼ö·Ïó(Citation) VOL 35 NO. 03 PP. 0054 ~ 0065 (2019. 12)
Çѱ۳»¿ë
(Korean Abstract)
´Ùº¯·® ½Ã°è¿­ µ¥ÀÌÅÍ´Â ±â»ó, ±³Åë, Á¦Á¶ µî ´Ù¾çÇÑ »ê¾÷¿¡¼­ ¿¹Ãø ¹× ÀÌ»ó ŽÁö µîÀÇ ¿©·¯ ÀÀ¿ëÀ» À§ÇØ È°¹ßÈ÷ È°¿ëµÈ´Ù. ÇÏÁö¸¸ ¿©·¯ ¿øÀο¡ ÀÇÇØ ¹ß»ýÇÏ´Â ´Ùº¯·® ½Ã°è¿­ µ¥ÀÌÅÍ ³»ÀÇ °áÃøÄ¡´Â µ¥ÀÌÅÍÀÇ Ç°Áú°ú È°¿ëµµ¸¦ Å©°Ô ¶³¾î¶ß¸°´Ù. µû¶ó¼­ ´Ùº¯·® ½Ã°è¿­ µ¥ÀÌÅÍ ³»¿¡¼­ÀÇ °áÃøÄ¡¸¦ ó¸®Çϱâ À§ÇÑ ´Ù¾çÇÑ ¿¬±¸°¡ ½ÃµµµÇ¾ú´Ù. ÇÏÁö¸¸ ´ëºÎºÐÀÇ ±âÁ¸ ¹æ¹ýµéÀº ´Ùº¯·® ½Ã°è¿­ µ¥ÀÌÅÍÀÇ °¡Àå Áß¿äÇÑ µÎ °¡Áö Ư¡ÀÎ º¯¼ö °£ »ó°ü°ü°è¿Í ½Ã°£»óÀÇ ÀÇÁ¸°ü°è¸¦ Á¦´ë·Î ¹Ý¿µÇÏÁö ¸øÇÏ¿´´Ù. º» ³í¹®¿¡¼­´Â À§ µÎ °¡Áö Ư¡À» È¿°úÀûÀ¸·Î °í·ÁÇÏ¿© °áÃøÄ¡ ó¸® ¼º´ÉÀ» Å©°Ô Çâ»óÇÑ µö·¯´× ±â¹ÝÀÇ ÃÖ±Ù ¿¬±¸µéÀ» ¼Ò°³ÇÑ´Ù. Á¦½ÃµÈ ¸ðµ¨µéÀº Å©°Ô È®Á¤Àû ¸ðµ¨°ú »ý¼º ¸ðµ¨·Î ±¸ºÐµÈ´Ù. º» ³í¹®Àº °¢ ¸ðµ¨ÀÇ Æ¯Â¡À» ±â¹ÝÀ¸·Î ÀÛµ¿¹æ½Ä°ú Àå´ÜÁ¡À» ÀÚ¼¼ÇÏ°Ô ¼³¸íÇÑ´Ù. ±×¸®°í µö·¯´×À» ÀÌ¿ëÇÑ ¹æ¹ýµéÀÇ ÇÑ°èÁ¡À» Áö¸ñÇϸç ÇâÈÄ ¿¬±¸ ¹æÇâ¿¡ ´ëÇÏ¿© ÅäÀÇÇÑ´Ù.
¿µ¹®³»¿ë
(English Abstract)
Multivariate time series data has been widely used for various applications (e.g. prediction and outlier detection) across many industries such as meteorology, transportation, and manufacturing. However, missing values in multivariate time series data significantly reduce the usability of the data. There have been many efforts to overcome this problem, but most of the previous studies do not effectively consider the two most important characteristics of multivariate time series data: (1) the correlation between variables and (2) the temporal dependency. This paper introduces the recent deep learning-based approaches which show better performances on processing missing values by taking into account these two characteristics. According to the mechanism of the model, the introduced studies can be categorized into either a discriminative model or a generative model. This paper explains the basic mechanism of each model, as well as its strengths and weaknesses. The limitations of the models and possible future research directions are also discussed.
Å°¿öµå(Keyword) µö·¯´×   µ¥ÀÌÅÍ ¸¶ÀÌ´×   ´Ùº¯·® ½Ã°è¿­ µ¥ÀÌÅÍ   °áÃøÄ¡   Deep Learning   Data Minning   Multivariate Time Series   Missing value  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå