µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
ÁýÇÕ¿¡ ±â¹ÝÇÑ ¼ºê½ÃÄö½º ¸ÅĪ ±â¹ý |
¿µ¹®Á¦¸ñ(English Title) |
A Technique for Set-based Subsequence Matching |
ÀúÀÚ(Author) |
¿©ÀºÁö
ÀÌÁÖ¿ø
ÀÓÈ¿»ó
Eunji Yeo
Juwon Lee
Hyo-Sang Lim
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 32 NO. 03 PP. 0152 ~ 0169 (2016. 12) |
Çѱ۳»¿ë (Korean Abstract) |
º» ³í¹®¿¡¼´Â ÁýÇÕ¿¡ ±â¹ÝÇÑ µ¥ÀÌÅͽºÆ®¸²¿¡¼ÀÇ ¼ºê½ÃÄö½º ¸ÅĪ ¹æ¹ýÀÎ S-Match(Set-based subsequence Matching)¸¦ Á¦¾ÈÇÏ¿´´Ù. ¼ºê½ÃÄö½º ¸ÅĪÀº µ¥ÀÌÅÍ ½ÃÄö½º(data sequence) Áß¿¡¼ ÁúÀÇ ½ÃÄö½º(query sequence)¿Í À¯»çÇÑ ¼ºê½ÃÄö½º¿Í ÇØ´ç ¼ºê½ÃÄö½ºÀÇ À§Ä¡¸¦ ã´Â ¹®Á¦ÀÌ´Ù. S-Match´Â ´ÙÀ½ÀÇ µÎ °¡Áö Ư¡À» °®´Â´Ù. ù ¹ø°·Î »ç¿ëÀÚÀÇ ¼±È£¸¦ ÁýÇÕ °³³äÀ» °í·ÁÇÏ¿© ¡°¼±È£ ¾ÆÀÌÅÛ ÁýÇÕ ½ÃÄö½º¡±·Î Ç¥ÇöÇÏ¿© ½Ã°£ °³³äÀ» °í·ÁÇϸ鼵µ Á¤È®ÇÑ ¼ø¼¿¡ ÀÇÇÑ ºÒÀÏÄ¡ ¹®Á¦¸¦ ÇØ°áÇÏ¿´´Ù. À̶§ ¾ÆÀÌÅÛ ÁýÇÕ ½ÃÄö½º °£ÀÇ À¯»çµµ¸¦ ÃøÁ¤Çϱâ À§ÇØ À¯Å¬¸®µð¾È °Å¸®¸¦ ÁýÇÕÀ¸·Î È®ÀåÇÑ À¯Å¬¸®µð¾È ÁýÇÕ °Å¸®¸¦ Á¦¾ÈÇÏ¿´´Ù. µÎ ¹ø°·Î Ãßõ ½Ã½ºÅÛ(Recommendation System)ÀÇ ÇÙ½É ¿ä¼ÒÀÎ À¯»ç »ç¿ëÀÚ ¸ÅĪ ¹®Á¦¸¦ µ¥ÀÌÅͽºÆ®¸²¿¡¼ÀÇ ¼ºê½ÃÄö½º ¸ÅĪ ¹®Á¦·Î º¯È¯ÇÏ¿© ´Ù¸¥ »ç¿ëÀÚÀÇ ÃÖ±Ù ¼±È£»Ó¸¸ ¾Æ´Ï¶ó °ú°ÅÀÇ ¸ðµç ½ÃÁ¡ÀÇ ¼±È£±îÁöµµ °Ë»öÇÏ¿´´Ù. ±×¸®°í S-Match¸¦ ¼öÇàÇÒ ¶§¿¡ ½ÇÁ¦·Î À¯»çÇÏÁö¸¸ À¯»çÇÏÁö ¾Ê´Ù°í ÆǴܵǴ Âø¿À±â°¢ÀÌ ¹ß»ýÇÏÁö ¾ÊÀ½À» Áõ¸íÇÏ¿´´Ù. ¼º´É Æò°¡ °á°ú, Á¦¾ÈÇÏ´Â S-Match°¡ ½ÇÁ¦ ¿µÈ ÆòÁ¡ µ¥ÀÌÅÍ¿¡¼ ¼ºê½ÃÄö½º ¸ÅĪÀ» ¼öÇàÇÏ¿© Âø¿À±â°¢ÀÌ ¾øÀÌ Á¤È®ÇÏ°Ô À¯»çÇÑ »ç¿ëÀÚ¸¦ ã¾Æ³»´Â °ÍÀ» º¸¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
In this paper, we propose a method for set-based subsequence matching (S-Match) in data streams. Subsequence matching is a problem to find subsequences and their locations in data sequences which are similar to a query sequence. We first propose the preferred item set sequence which reflects the time concept of user preference. A preferred item set sequence is an ordered list of sets where each set collects preferred items within in a specific time interval. We then propose a similarity measurement between item set sequences, the Euclidean set distance, which extends Euclidian distance. Second, in order to find the similar user not only in current time but also in past time, we transforms the similar user matching problem into the similar subsequence matching problem. We proves that the method does not incur false dismissals which are actually similar but discarded in the results of the similar sequence matching. Through experiments with movie rating real data sets, we show that S-Match accurately finds similar users with a false dismissal.
|
Å°¿öµå(Keyword) |
¼ºê½ÃÄö½º ¸ÅĪ
µ¥ÀÌÅͽºÆ®¸²
Ãßõ ½Ã½ºÅÛ
¾ÆÀÌÅÛ ÁýÇÕ
Subsequence matching
data stream
recommendation system
item set
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|