full2010.pdf - page 451

413
Optimization, are presented in this research. The completed stocks of Stock Exchange of Thailand are randomly
selected to test with three techniques. The success of this model is based on two performance measures, hit rate
and trained running time. According to the comparison among three techniques, Back Propagation Neural
Network is shown in a very good outcome for accurate prediction. Therefore, Back Propagation Neural Network
is selected to implement for Thai stock market assistant system. This system can assist new investors for the better
decision.
Keywords : Back propagation neural network, Stock Market Prediction, Stock Exchange of Thailand - SET
‡ÎµœÎµ
ÁœºÉ°Š‹µ„Äœž{‹‹»´ œ„µ¦¨Šš»œÄœ˜¨µ—®¨´„š¦´ ¡¥rœ´ Å—oªn µÁž}œš¸Éœ· ¥¤°¥nµŠ¤µ„ ˜n„µ¦¨Šš»œœ´Ê œ ¥n°¤¤¸
‡ªµ¤Á¸É¥Š˜nµŠÇ¤µ„¤µ¥ ×¥ÁŒ¡µ³œ´„¨Šš»œ¤º°Ä®¤n —oª¥Á®˜» œ¸Ê‹¹Š¤¸ „µ¦œÎµ„¨ª·›¸˜nµŠÇ¤µÄošÎµœµ¥ Á¡ºÉ°¨—‡ªµ¤Á¸É¥Š
š¸É ‹³Á„·— ¹Êœ„´œ´„¨Šš»œ „µ¦ª· Á‡¦µ³®r®»o œÃ—¥Äoª·›¸ šµŠÁš‡œ·‡ Ž¹É ŠÁž}œ„µ¦ª·Á‡¦µ³®r„µ¦Á‡¨ºÉ°œ˜´ª °Š¦µ‡µ®»oœÂ¨³
ž¦· ¤µ–„µ¦ŽºÊ° µ¥‹µ„ o°¤¼¨Äœ°—¸˜Á¡ºÉ°š¸É‹³‡µ—„µ¦–rœªÃœo ¤ °Š¦µ‡µÄœ°œµ‡˜ ×¥„µ¦ª·Á‡¦µ³®ršµŠÁš‡œ·‡
°¥n µŠŠnµ¥Á¦·É¤—oª¥„µ¦¡·‹µ¦–µÂœªÃœo¤„µ¦ ¹Êœ¨Š‹µ„„¦µ¢ °Š¦µ‡µ®»oœ ®¨´Š‹µ„œ´ÊœÅ—o¤¸„µ¦Äo Áš‡œ·‡„µ¦ª·Á‡¦µ³®r
šµŠ™·˜·¤µnª¥Äœ„µ¦¡¥µ„¦–r Ánœ White(1989) ŗočo˜´ªÂ ARIMA Ž¹É ŠÄo„´œ°¥nµŠÂ¡¦n®¨µ¥Äœ„µ¦¡¥µ„¦–r
o°¤¼¨°œ» „¦¤Áª¨µš¸ÉÁž}œ o°¤¼¨Á·ŠÁo œ ˜n°¥n µŠÅ¦„Șµ¤Áš‡œ· ‡—´Š„¨nµª¥´ŠÅ¤nÁ®¤µ³„´„µ¦¡¥µ„¦–r ®»oœŽ¹ÉŠÁž}œ o°¤¼¨
š¸É ŤnÁž}œÁ· ŠÁoœ ˜n°¤µ Li Sheng (2008) ª´‹œ£´ š¦(2551) Clarence(1993) Zekic(1998) ¨³¤œ˜¦¸(2551)ŗo¤¸„µ¦Äo
˜´ªÂÃ‡¦Š nµ¥ž¦³µšÁš¸ ¥¤Â˜nµŠÇ œÎ µ¤µÄo 宦´¡¥µ„¦–r®»o œŽ¹ÉŠ˜´ªÂœ¸Ê Áž}œ˜´ªÂš¸É¤¸‡ªµ¤µ¤µ¦™Äœ
„µ¦Á¦¸¥œ¦¼o‹µ„ o°¤¼¨Äœ°—¸ ˜Â¨oªœÎµÅžž¦´ ‡nµ™nªŠœÊ宜´„宦´¡¥µ„¦–r ˜n—oª¥Ã‡¦Š nµ¥ž¦³µšÁš¸¥¤¥´Š¤¸ž{ ®µ
Ĝ„µ¦Á¨º°„™µž{˜¥„¦¦¤ ‡º°Å¤n¦¼oªnµ‡ª¦Á¨º°„‹Î µœªœ´ÊœŽn°œÂ¨³‹Î µœªœÃ®œ—Äœ´ÊœŽn°œÁž}œÁšnµÄ—‹¹Š‹³šÎµÄ®o
„µ¦¡¥µ„¦–r¤¸‡ªµ¤Ÿ·—¡¨µ—œo°¥» — µŠŠµœª·‹´¥Å—o¤¸Âœª‡·—Äœ„µ¦Á¡·É¤‹ÎµœªœÃ®œ— White (1989) ×¥Á¦·É ¤˜oœ‹µ„
‹ÎµœªœÃ®œ—Äœ´ÊœŽn°œš¸Éœo °¥„n°œ ¨o ª‹¹Š‡n°¥ÇÁ¡·É¤‹ÎµœªœÃ®œ—‹œÅ—o‡nµš¸É¡°Ä‹ ĜµŠŠµœª· ‹´¥Å—o¤¸„µ¦Äo
ª·›¸„µ¦™—™°¥Á· ŠÁoœÄœ„µ¦¨—‹ÎµœªœÃ®œ—Äœ´ÊœŽn°œ Fletcher (1998) ×¥Á¦·É¤Ä®oǦŠ nµ¥Á¦¸¥œ¦¼o —oª¥‹Îµœªœ
‹Îµœªœ®œ¹ÉŠÂ¨oªš—¨°Š¨—‹ÎµœªœÃ®œ—¨Š®¨´ Š‹µ„œ´ÊœÄo ™·˜·š—°ªnµ‡ª¦¨—®¦º°Å¤n ‹µ„ oµŠ˜oœ¡ªn µ
Šµœª·‹´¥nªœÄ®n¥´Š‡ŠÁž}œ„µ¦‡oœ®µ ‡Îµ˜°š¸É‹³šÎ µÄ®o„µ¦¡¥µ„¦–r¤¸‡ªµ¤Ÿ·—¡¨µ—œo°¥»—¨³Á¡·É¤ž¦³·š›·£µ¡
Ĝ„µ¦¡¥µ„¦–r®»o œ Šµœª·‹´¥œ¸Ê‹¹ŠÅ—ošÎ µ„µ¦š—¨°ŠÃ—¥Á¨º°„čo µ¤˜´ªÂ‡º° Back Propagation Neural Network,
Fuzzy-Back Propagation Neural Network ¨³ Fuzzy-Particle Swarm Optimization Neural Network Á¡ºÉ°š¸É‹³—¼ªnµ
˜´ªÂÅ®œ‹³Áž}œ˜´ªÂš¸É Á®¤µ³¤š¸É »— ×¥šÎµ„µ¦š—¨°ŠÁž¦¸¥Áš¸ ¥ ǦŠ nµ¥ž¦³µšÁš¸¥¤ „n°œ™¼ „
ž¦´ž¦»Š ¨³ ž¦´ ž¦»ŠÂ¨o ª Á¡ºÉ°®µ ǦŠ nµ¥ž¦³µšÁš¸ ¥¤š¸ÉÁ®¤µ³¤„´ ®»oœÄœ˜¨µ—®¨´„š¦´¡¥r®nŠž¦³Áš«Åš¥
Ž¹ÉŠ‹³šÎµ„µ¦»n ¤®»oœÄœ˜¨µ—®¨´ „š¦´¡¥r®nŠž¦³Áš«Åš¥ 2 že¥o°œ®¨´Š (2551-2552) ¤µš—°„´š´ÊŠµ¤˜´ªÂ
‹µ„„µ¦š—°„´˜´ ªÂ˜nµŠÇ ¡ªnµ ‡n µ‡ªµ¤Á¦Èªš¸Éčo Ĝ„µ¦Á¦¸¥œ¦¼o ¨³‡ªµ¤µ¤µ¦™Äœ„µ¦šÎ µœµ¥ °Š Back
Propagation Neural Network Á®¤µ³¤„´„µ¦¡¥µ„¦–r¦µ‡µžd— °Š®»oœ¦³¥³´Ê œÄœ˜¨µ—®¨´„š¦´¡¥r®nŠž¦³Áš«
Ś¥ Î µ®¦´„µ¦ž¦³Á¤·œŸ¨„µ¦¡¥µ„¦–r°Š˜´ªÂœ´ Áž}œ·ÉŠÎµ‡´ Á¡¦µ³®µ„˜o°Š„µ¦ª´ —ªnµ˜´ªÂ¡¥µ„¦–rŗo —¸
¤µ„œo °¥Á¡¸¥ŠÄ—‹³˜o °Š¤¸˜´ªÂš¸Éčo 宦´°oµŠ°·ŠÅ—oŠµœª·‹´ ¥œ¸Ê‹¹ŠÅ—ož¦³Á¤·œŸ¨ °Š˜´ ªÂ¡¥µ„¦–r®»o œ—oª¥„µ¦