µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
µö·¯´× ¾Ó»óºíÀ» ÀÌ¿ëÇÑ ÁÖ°¡¿¹Ãø |
¿µ¹®Á¦¸ñ(English Title) |
Stock Price Prediction Using Deep Learning Ensemble |
ÀúÀÚ(Author) |
±èÈ«Áö
Á¤ÁöÇö
°íÀº³ª·¡
Á¶¸¸Àç
À̱âÈÆ
Hong-Ji Kim
Ji-Hyun Jung
Eun-Na-Rae Ko
Man-Jae Cho
Ki-Hoon Lee
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 34 NO. 02 PP. 0111 ~ 0120 (2018. 08) |
Çѱ۳»¿ë (Korean Abstract) |
ÃÖ±Ù µö·¯´×(Deep Learning)À» ÀÌ¿ëÇÑ ÁÖ°¡¿¹ÃøÀÌ È°¹ßÇÏ°Ô ¿¬±¸µÇ°í ÀÖÀ¸³ª, ¼·Î ´Ù¸¥ µö·¯´× ¸ðµ¨µéÀ» °áÇÕÇÏ´Â ¾Ó»óºí(Ensemble) ¹æ¹ý¿¡ ´ëÇÑ ¿¬±¸´Â Ãʱ⠴ܰèÀÌ´Ù. µö·¯´× ¸ðµ¨¿¡´Â Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN)ÀÌ ÀÖ´Ù. º» ³í¹®¿¡¼´Â ¼¼ °¡Áö µö·¯´× ¸ðµ¨(MLP, CNN, RNN)ÀÌ ¿¹ÃøÇÑ °á°ú¸¦ °áÇÕÇÏ°í MLP¸¦ »ç¿ëÇÏ¿© ´Ù½Ã ÇнÀÇÏ´Â ½ºÅÂÅ·(Stacking) ±â¹ÝÀÇ ¾Ó»óºí ¸ðµ¨À» »ç¿ëÇÏ¿© ÁÖ°¡¸¦ ¿¹ÃøÇÑ´Ù. KOSPI »óÀ§ 30 Á¾¸ñ Áß 18°³ Á¾¸ñÀ» ÀÌ¿ëÇÏ¿© ½ÇÇèÇÑ °á°ú, Á¦¾ÈÇÑ ¹æ¹ýÀÌ ±âÁ¸ ¹æ¹ý¿¡ ºñÇØ Àý´ëÆò±Õ¹éºÐÀ²¿ÀÂ÷(MAPE)°¡ 8.74%¿¡¼ 3.35%·Î °¨¼ÒÇÏ¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
Recently, there have been research efforts on predicting stock price using deep learning, but little attention has been paid so far to ensemble methods, which combines different deep learning models. Deep learning models include Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), and Recurrent Neural Network (RNN). In this paper, we propose a stacking-based ensemble model where a deep learning model combines predictions of three different deep learning models (MLP, CNN, and RNN). We use MLP as the second level model. The experimental results using 18 stock items among KOSPI top 30 items show that the proposed method improves the mean absolute percentage error (MAPE) from 8.74%, which is the MAPE of the state-of-the-art method, to 3.35%.
|
Å°¿öµå(Keyword) |
µö·¯´×
¾Ó»óºí
½ºÅÂÅ·
ÁÖ°¡¿¹Ãø
Deep Learning
Ensemble
Stacking
Stock Price Pre
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|