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

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ³í¹®Áö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ³í¹®Áö > Çѱ¹Á¤º¸°úÇÐȸ ³í¹®Áö > Á¤º¸°úÇÐȸ ³í¹®Áö D : µ¥ÀÌŸº£À̽º

Á¤º¸°úÇÐȸ ³í¹®Áö D : µ¥ÀÌŸº£À̽º

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) ¿Àµð¼¼¿ì½º/DM: ¿Àµð¼¼¿ì½º °´Ã¼ °ü°èÇü DBMS¿Í ¹Ð°áÇÕµÈ µ¥ÀÌÅÍ ¸¶ÀÌ´× ½Ã½ºÅÛÀÇ ¼³°è ¹× ±¸Çö
¿µ¹®Á¦¸ñ(English Title) ODYSSEUS/DM: Design and Implementation of a Data Mining System Tightly-Coupled with the Odysseus Object-Relational DBMS
ÀúÀÚ(Author) ÀÌÀÏ¿±   ±è¹Î¼ö   ±èÁؼº   ÀÌÁ¤ÈÆ   Ȳ±Ô¿µ   Il-Yeop Yi   Min Soo Kim   Jun-Sung Kim   Jeong-Hoon Lee   Kyu-Young Whang  
¿ø¹®¼ö·Ïó(Citation) VOL 39 NO. 04 PP. 0211 ~ 0220 (2012. 08)
Çѱ۳»¿ë
(Korean Abstract)
µ¥ÀÌÅÍÀÇ ¾çÀÌ ±ÞÁõÇÔ¿¡ µû¶ó µ¥ÀÌÅÍ ¸¶ÀÌ´×ÀÇ ´ë»óÀÎ µ¥ÀÌÅÍ°¡ ´ëºÎºÐ DBMS¿¡ ÀÇÇØ °ü¸®µÇ¸é¼­, µ¥ÀÌÅÍ ¸¶ÀÌ´× ±â´ÉÀ» DBMS¿Í °áÇÕÇÏ¿© ¼öÇàÇÏ´Â µ¥ÀÌÅÍ ¸¶ÀÌ´× ½Ã½ºÅÛÀÇ ¼º´ÉÀÌ Áß¿äÇÑ À̽´°¡ µÇ°í ÀÖ´Ù. µ¥ÀÌÅÍ ¸¶ÀÌ´× ±â´É°ú DBMSÀÇ °áÇÕ ¹æ¹ýÀº µ¥ÀÌÅÍ ¸¶ÀÌ´× ½Ã½ºÅÛÀÇ ¼º´É¿¡ Å« ¿µÇâÀ» ¹ÌÄ¡´Âµ¥, ¼Ò°áÇÕ°ú ¹Ð°áÇÕÀ¸·Î ±¸ºÐÇÒ ¼ö ÀÖ´Ù. µ¥ÀÌÅÍ ¸¶ÀÌ´× ±â´ÉÀ» DBMS ¿ÜºÎ¿¡ ±¸ÇöÇÏ´Â ¼Ò°áÇÕ ¹æ¹ýÀº DBMSÀÇ »óÀ§ ·¹º§ ÀÎÅÍÆäÀ̽º¸¦ »ç¿ëÇϱ⠶§¹®¿¡ ±¸ÇöÀº ¿ëÀÌÇÏÁö¸¸ ³ôÀº ¼º´ÉÀ» ±â´ëÇϱ⠾î·Æ°í, µ¥ÀÌÅÍ ¸¶ÀÌ´× ±â´ÉÀ» DBMS ³»ºÎ¿¡ ±¸ÇöÇÏ´Â ¹Ð°áÇÕ ¹æ¹ýÀº DBMSÀÇ ÇÏÀ§ ·¹º§ ÀÎÅÍÆäÀ̽º¸¦ »ç¿ëÇϱ⠶§¹®¿¡ ±¸ÇöÀÌ ¾î·ÆÁö¸¸ ³ôÀº ¼º´ÉÀ» º¸ÀδÙ. µ¥ÀÌÅÍ ¸¶ÀÌ´× ½Ã½ºÅÛÀÇ ¼º´É Çâ»óÀ» À§Çؼ­´Â ¹Ð°áÇÕ ¹æ¹ýÀÌ ÇʼöÀûÀÌÁö¸¸, ´ëºÎºÐÀÇ ±âÁ¸ ½Ã½ºÅÛµéÀº ¼Ò°áÇÕ ¹æ¹ýÀ¸·Î ±¸ÇöµÇ¾î ÀÖ´Ù. µû¶ó¼­, º» ³í¹®¿¡¼­´Â ¿ì¼öÇÑ ¼º´ÉÀÇ µ¥ÀÌÅÍ ¸¶ÀÌ´× ½Ã½ºÅÛ ±¸ÇöÀ» À§ÇØ µ¥ÀÌÅÍ ¸¶ÀÌ´× ±â´ÉÀ» ¿Àµð¼¼¿ì½º °´Ã¼ °ü°èÇü DBMS¿¡ ¹Ð°áÇÕÇÏ´Â ¹æ¹ýÀ» Á¦¾ÈÇÏ°í, À̸¦ µû¸£´Â ¿Àµð¼¼¿ì½º/DM(Data Mining)À» ±¸ÇöÇÑ´Ù. ¿Àµð¼¼¿ì½º/DMÀº µ¥ÀÌÅÍ ¸¶ÀÌ´× ±â´É ¼öÇà¿¡ ÇÊ¿äÇÑ Å¸ÀÔ ¹× ¿¬»êÀ» DBMS ¿£Áø ³»ºÎ¿¡ ±¸ÇöÇÏ¿© ¼Ò°áÇÕ ¹æ¹ý¿¡¼­ ¹ß»ýÇÏ´Â ¿À¹öÇìµå¸¦ ÃÖ¼ÒÈ­ÇÑ´Ù. ¶ÇÇÑ, ¿Àµð¼¼¿ì½º/DMÀº ±âÁ¸ ½Ã½ºÅÛµé°úÀÇ »óÈ£ ¿î¿ë¼ºÀ» ³ôÀ̱â À§ÇØ µ¥ÀÌÅÍ ¸¶ÀÌ´× °á°ú Ç¥ÇöÀ» À§ÇÑ »ê¾÷ Ç¥ÁØÀÎ PMML(Predictive Model Markup Language)À» Áö¿øÇÑ´Ù. ½ÇÇè¿¡¼­´Â ¿Àµð¼¼¿ì½º/DMÀÇ ¼º´ÉÀÌ ¼Ò°áÇÕ ¹æ¹ýÀ¸·Î ±¸ÇöÇÑ µ¥ÀÌÅÍ ¸¶ÀÌ´× ½Ã½ºÅÛ¿¡ ºñÇØ ¿ì¼öÇÔÀ» º¸ÀδÙ.
¿µ¹®³»¿ë
(English Abstract)
As the amount of data increases rapidly, the data to be mined is more and more managed in the DBMS. Thus, the performance of data mining systems integrating data mining functionalities with the DBMS becomes an important issue. The methods of integrating data mining functionalities with the DBMS have an effect on the performance of the data mining system and can be classified into two categories: loose-coupling and tight-coupling. In loose-coupling, data mining functionalities are implemented outside the DBMS. Since it uses a high-level interface of the DBMS, it is easy to implement, but difficult to achieve high performance. In tight-coupling, data mining functionalities are implemented inside the DBMS. Since it uses a low-level interface of the DBMS, its implementation is non-trivial but easy to achieve high performance. Although the implementation of a high-performance data mining system requires the tight-coupling architecture, most of the existing systems are implemented in the loose-coupling architecture. Therefore, in order to implement a high performance data mining system, we introduce an architecture that tightly couples data mining functionalities with the Odysseus object-relational DBMS, and implement Odysseus/DM (Data Mining) that adopts this architecture. By implementing a type and operations related to data mining functionalities inside the DBMS engine, Odysseus/DM minimizes the overhead that can be incurred in the loose-coupling architecture. Odysseus/DM supports PMML (Predictive Model Markup Language), which is an industrial standard for representing data mining results, in order to provide interoperability with the existing systems. Finally, we conduct experiments showing that the performance of Odysseus/DM is superior to that of a data mining system implemented in the loose-coupling architecture.
Å°¿öµå(Keyword) µ¥ÀÌÅÍ ¸¶ÀÌ´×   ¹Ð°áÇÕ   Data mining   DBMS   Tight-coupling  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå