• 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 B : ¼ÒÇÁÆ®¿þ¾î ¹× ÀÀ¿ë

Á¤º¸°úÇÐȸ ³í¹®Áö B : ¼ÒÇÁÆ®¿þ¾î ¹× ÀÀ¿ë

Current Result Document : 16 / 44 ÀÌÀü°Ç ÀÌÀü°Ç   ´ÙÀ½°Ç ´ÙÀ½°Ç

ÇѱÛÁ¦¸ñ(Korean Title) ´ÙÁß ·¹ÀÌºí µ¥ÀÌÅÍ ºÐ·ù¸¦ À§ÇÑ »óÈ£ Á¤º¸ ôµµ¸¦ ÀÌ¿ëÇÑ Æ¯Â¡ ¼±º° ±â¹ý
¿µ¹®Á¦¸ñ(English Title) Using Mutual Information for Selecting Features in Multi-label Classification
ÀúÀÚ(Author) ÀÓÇö±â   ±è´ë¿ø   Hyunki Lim   Dae-Won Kim  
¿ø¹®¼ö·Ïó(Citation) VOL 39 NO. 10 PP. 0806 ~ 0811 (2012. 10)
Çѱ۳»¿ë
(Korean Abstract)
ÃÖ±Ù ¸¹Àº ÀÀ¿ë¿¡¼­ ´ÙÁß ·¹ÀÌºí µ¥ÀÌÅÍ°¡ ¹ß»ýÇÏ°í ÀÖ´Ù. ÇÏÁö¸¸ ÀÌ µ¥ÀÌÅÍ´Â ±âÁ¸ ±â°è ÇнÀ, µ¥ÀÌÅÍ ¸¶ÀÌ´× ºÐ¾ßÀÇ ¹æ¹ý Àû¿ëÀÌ ¾î·Æ´Ù. ±× ÀÌÀ¯´Â Å©°Ô µÎ °¡Áö·Î ±âÁ¸ ¹æ¹ýµéÀÌ ´ÜÀÏ ·¹ÀÌºí µ¥ÀÌÅÍ¿¡ ÃÊÁ¡À» ¸ÂÃß°í ÀÖ´Ù´Â °Í°ú ´ÙÁß ·¹ÀÌºí µ¥ÀÌÅÍÀÇ Æ¯¼ºÀ» ¹Ý¿µÇÏÁö ¸øÇÏ°í ÀÖ´Ù´Â °ÍÀÌ´Ù. ´ëºÎºÐÀÇ Æ¯Â¡ ¼±º° ±â¹ýÀº ´ÜÀÏ ·¹ÀÌºí µ¥ÀÌÅÍ¿¡ ÃÊÁ¡À» ¸ÂÃß°í Àֱ⠶§¹®¿¡ ´ÙÁß ·¹ÀÌºí µ¥ÀÌÅÍ¿¡´Â ±âÁ¸ Ư¡ ¼±º° ±â¹ýµéÀ» Àû¿ëÇÒ ¼ö ¾ø´Ù. ´ÙÁß ·¹ÀÌºí µ¥ÀÌÅÍ¿¡ Ư¡ ¼±º° ±â¹ýÀ» Àû¿ëÇϱâ À§Çؼ­ ´ÙÁß ·¹ÀÌºí µ¥ÀÌÅ͸¦ ´ÜÀÏ ·¹ÀÌºí µ¥ÀÌÅÍ·Î ÀüȯÇÏ´Â ¹æ¹ýµéÀÌ »ç¿ëµÈ´Ù. ÇÏÁö¸¸ ·¹ÀÌºí º¯È¯Àº µ¥ÀÌÅÍ °íÀ¯ÀÇ Æ¯¼ºÀ» ¹Ý¿µÇÏÁö ¸øÇÏ°í Á¤º¸ ¼Õ½ÇÀ» °¡Á®¿Ã ¼ö ÀÖ´Ù. º» ³í¹®Àº ·¹À̺í°ú ·¹ÀÌºí »çÀÌÀÇ ¿¬°ü¼ºÀ» °í·ÁÇÏ¿© ´ÙÁß ·¹ÀÌºí µ¥ÀÌÅÍ¿¡ ¹Ù·Î Àû¿ëÇÒ ¼ö Àִ Ư¡ ¼±º° ±â¹ýÀ» Á¦¾ÈÇÑ´Ù. Á¦¾ÈÇÏ´Â ¹æ¹ýÀÇ ¿ì¼ö¼ºÀ» º¸À̱â À§ÇØ Å¬·¡½º ºÐ·ù ½ÇÇèÀ» ÇÏ¿´´Ù. À̸¦ ÅëÇØ ±âÁ¸ Ư¡ ¼±º° ±â¹ýµé¿¡ ºñÇؼ­ Á¦¾ÈÇÏ´Â ±â¹ýÀÇ ¼º´ÉÀÌ ¿ì¼öÇÏ´Ù´Â °ÍÀ» º¸¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Lately multi-label data set occurs in many applications. However it is difficult to apply in machine learning and data mining fields. There are two reasons: One is that most of researches are focusing on the single-label problem and the other is that the previous methods do not account the characteristics of multi-label. Existing methods cannot be applied to multi-label data because most of feature selection methods have focused in single-label data. For applying existing method, there have been used label transformation methods. However label transformation may lead to information loss of data. In this paper, we propose feature selection method for multi-label data considering the dependency between labels. We experimented classification for demonstrating the superiority of proposed method. This shows that the proposed method is better than previous feature selection methods.
Å°¿öµå(Keyword) Ư¡ ¼±º° ±â¹ý   ´ÙÁß·¹ÀÌºí µ¥ÀÌÅÍ   Ŭ·¡½º ºÐ·ù   »óÈ£ Á¤º¸ ôµµ   Feature Selection   Multi-label Data   Classification   Mutual Information  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå