µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
¾Ó»óºí ºÐ·ù ¸ðµ¨À» ÀÌ¿ëÇÑ ½ºÅ¸Æ®¾÷ ¹Ì·¡°¡Ä¡ ¿¹Ãø |
¿µ¹®Á¦¸ñ(English Title) |
Startup Future Value Prediction Using Ensemble Classification Model |
ÀúÀÚ(Author) |
½ÅÈñ¹Î
±è»óÈ£
À̱âÈÆ
Heemin Shin
Sangho Kim
Ki-Hoon Lee
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 38 NO. 02 PP. 0042 ~ 0054 (2023. 01) |
Çѱ۳»¿ë (Korean Abstract) |
Çѱ¹ °æÁ¦°¡ Àú¼ºÀå´Ü°è¿¡ ÁøÀÔÇÏ¸é¼ ³ôÀº ¼ºÀå °¡´É¼ºÀ» °¡Áø ½ºÅ¸Æ®¾÷¿¡ ´ëÇÑ ÅõÀÚ°¡ Áõ°¡ÇÏ°í ÀÖ´Ù. ½ºÅ¸Æ®¾÷Àº »óÀå±â¾÷°ú ´Þ¸® ±â¾÷¿¡ ´ëÇÑ Á¤º¸°¡ ºÎÁ·Çϱ⠶§¹®¿¡ ÅõÀÚ¸¦ À§ÇÑ ¹Ì·¡°¡Ä¡¸¦ ¿¹ÃøÇÏ´Â °ÍÀÌ ¾î·Æ´Ù. º» ³í¹®¿¡¼´Â ¾Ó»óºí ºÐ·ù ¸ðµ¨À» ÀÌ¿ëÇØ ½ºÅ¸Æ®¾÷ÀÇ ¹Ì·¡°¡Ä¡¸¦ ¿¹ÃøÇÏ´Â ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. ÇöÀç °¡Ä¡¿Í ºñ±³ÇÑ ¹Ì·¡°¡Ä¡ÀÇ º¯ÈÀ²À» ±âÁØÀ¸·Î ¸î °³ÀÇ Å¬·¡½º·Î ³ª´©°í ÁÖ¾îÁø ½ºÅ¸Æ®¾÷ÀÌ ¾î´À Ŭ·¡½º¿¡ ¼ÓÇÒÁö¸¦ ¿¹ÃøÇÏ¿© ÅõÀÚÀÇ»ç°áÁ¤¿¡ µµ¿òÀ» ÁØ´Ù. Á¦¾ÈÇÑ ¹æ¹ýÀº ºÐ·ù ¸ðµ¨ ÇнÀÀ» À§ÇØ ½ºÅ¸Æ®¾÷ÀÇ À繫, ºñÀ繫, ±â¾÷°¡Ä¡ µ¥ÀÌÅ͸¦ ¼öÁýÇÑ´Ù. ±×¸®°í 18°³ÀÇ ºÐ·ù ¸ðµ¨ Áß Á¤È®µµ°¡ °¡Àå ³ôÀº 5°³ÀÇ ¸ðµ¨À» °áÇÕÇÏ´Â ¼ÒÇÁÆ® ÅõÇ¥ ±â¹ÝÀÇ ¾Ó»óºí(ensemble) ¹æ¹ýÀ» »ç¿ëÇÏ¿© ½ºÅ¸Æ®¾÷ÀÇ Å¬·¡½º¸¦ ¿¹ÃøÇÑ´Ù. ½ÇÁ¦ ½ºÅ¸Æ®¾÷ ÅõÀÚ µ¥ÀÌÅÍ 206°³¸¦ ÀÌ¿ëÇÏ¿© ½ÇÇèÇÑ °á°ú, Á¦¾ÈÇÑ ¹æ¹ýÀÌ ÁÖ°¡¸ÅÃâ¾×ºñÀ²(PSR) ±â¹ÝÀÇ »ó´ëÀû °¡Ä¡Æò°¡ ¹æ¹ýº¸´Ù Æò±ÕÀûÀ¸·Î 17%P ´õ ³ôÀº ºÐ·ù Á¤È®µµ¸¦ º¸¿´´Ù. |
¿µ¹®³»¿ë (English Abstract) |
As the Korean economy enters a low-growth phase, investments in startups with high growth potential are increasing. Unlike listed companies, it is difficult to predict the future value of a startup company for investment because there is a lack of information on the startup. In this paper, we propose a method that predicts the future value of startups using an ensemble classification model. To help investment decisions, we divide startups into a few classes based on the changing rate of the future value compared with the current value and predict which class a given startup will belong to. The proposed method collects financial, non-financial, and company value data from startups for training classification models. It predicts the class of the startup using a soft-voting ensemble method that combines the five most accurate models among 18 classification models. The experimental results using 206 startup investment data show that the proposed method accomplishes 17%P higher classification accuracy on average than the relative valuation method based on Price Sales Ratio (PSR). |
Å°¿öµå(Keyword) |
½ºÅ¸Æ®¾÷
±â¾÷°¡Ä¡ ¿¹Ãø
¸Ó½Å·¯´×
ºÐ·ù ¸ðµ¨
¾Ó»óºí
startup
company value prediction
machine learning
classification model
ensemble
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|