• 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 ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) µö·¯´×À» À§ÇÑ ½ÇÇèÀû Æò°¡ ±â¹ÝÀÇ À̵¿°æ·Î µ¥ÀÌÅÍ ÀÎÄÚµù ¹æ¹ý ºñ±³
¿µ¹®Á¦¸ñ(English Title) Analysis of Trajectory Encoding Methodology Analysis Based on Experimental Evaluation for Deep Learning
ÀúÀÚ(Author) °­ÁØÇõ   ±è¹Î¼®   ÀÌÀç±æ   Junhyeok Kang   Minseok Kim   Jae-Gil Lee  
¿ø¹®¼ö·Ïó(Citation) VOL 25 NO. 08 PP. 0402 ~ 0406 (2019. 08)
Çѱ۳»¿ë
(Korean Abstract)
¼¾¼­±â¼úÀÇ ¹ß´Þ·Î ÀÚµ¿Â÷³ª »ç¶÷À¸·ÎºÎÅÍ ¼öÁýµÇ´Â À̵¿°æ·Î µ¥ÀÌÅÍÀÇ ¾çÀÌ ±ÞÁõÇÏ°í ÀÖ´Ù. ÀÌ·¯ÇÑ À̵¿°æ·Î µ¥ÀÌÅͷκÎÅÍ À¯ÀǹÌÇÑ °á°ú¸¦ À̲ø¾î³»±â À§Çؼ­ ÇöÀç±îÁö ¸¹Àº ¿¬±¸µéÀÌ È°¹ßÈ÷ ÁøÇàµÇ°í ÀÖ´Ù. ±× Áß ÀÚ¿¬¾î ó¸® µî Ÿ ºÐ¾ß¿¡¼­ µÎ°¢À» ³ªÅ¸³½ µö·¯´×À» ÀÌ¿ëÇÑ ¿¬±¸µéÀÌ ÃÖ±Ù ½ÃµµµÇ¾úÀ¸³ª, À̵¿°æ·Î µ¥ÀÌÅ͸¸ÀÇ °íÀ¯ÇÑ Æ¯¼ºÀ» È¿°úÀûÀ¸·Î ÀÎÄÚµùÇÏ´Â ¹æ¹ýÀº ¾ÆÁ÷ ¸¹ÀÌ ¿¬±¸µÇÁö ¾Ê¾Ò´Ù. º» ³í¹®¿¡¼­´Â À̵¿°æ·Î µ¥ÀÌÅÍÀÇ Æ¯¼ºÀ» Ç¥ÇöÇÏ´Â ÀÎÄÚµù ¹æ¹ý·Ð 5°¡Áö¿Í ±×¿¡ ¸Â´Â Á¤±ÔÈ­°úÁ¤À» ¼Ò°³ÇÏ°í, ÀÌ¿¡ ´ëÇÑ ºñ±³½ÇÇèÀ» ÅëÇØ ¼º´ÉÀ» ºñ±³ÇÏ°íÀÚ ÇÑ´Ù. Microsoft Geolife µ¥ÀÌÅ͸¦ È°¿ëÇÏ¿© LSTM±â¹ÝÀÇ µö·¯´× ¸ðµ¨À» ÇнÀÇÏ°í ±³Åë¼ö´Ü(Transportation Mode) ºÐ·ù¸¦ ¼öÇàÇÑ °á°ú, ÀÓº£µù°ú °áÇÕµÈ One-Hot Encoding ±â¹Ý Á¤±ÔÈ­ ¹æ¹ý·ÐÀÌ ¼º´É°ú ¼Óµµ Ãø¸é¿¡¼­ ºñ±³¸ðµ¨ Áß °¡Àå ¿ì¼öÇÑ °ÍÀ¸·Î °üÃøµÇ¾ú´Ù.
¿µ¹®³»¿ë
(English Abstract)
The amount of trajectory data collected from humans and vehicles is growing rapidly owing to the improvement in sensor-based technologies. Many researchers have actively attempted to draw meaningful insights from trajectory data. Recently, deep learning-based methods, which demonstrated high performance in other fields such as natural language processing have been used to tackle this issue. However, there is little effort on the development of effective methods to encode trajectory data. This paper introduces five trajectory encoding methods along with their normalization and compares their performance through an extensive evaluation. In the experiments on LSTM-based transportation mode classification using Microsoft Geolife data, merging of embedding and one-hot encoding along with normalization demonstrated the highest performance in terms of performance and model convergence speed.
Å°¿öµå(Keyword) À̵¿°æ·Î   µ¥ÀÌÅÍ ¸¶ÀÌ´×   µö·¯´×   ÀÎÄÚµù ¹æ¹ý·Ð   trajectory   data mining   deep learning   encoding methodology  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå