full2010.pdf - page 453

415
ž¦³Á£š®»oœ œªÃœo¤ °Š®»oœ ®¨´„Á„–”rĜ„µ¦‹ÎµÂœ„
ž¦³Á£šš¸É 1 ¹Êœ¤µ„ (¦µ‡µžd—ª´œ™´—Åž – ¦µ‡µžd —ª´œž{‹‹»´ œ)*100 / ¦µ‡µžd—ª´œž{‹‹»´ œ > 5
ž¦³Á£šš¸É 2 ¹ÊœÁ¨È„œo °¥ (¦µ‡µžd—ª´œ™´—Åž – ¦µ‡µžd —ª´œž{‹‹»´ œ)*100 / ¦µ‡µžd—ª´œž{‹‹»´ œ > 2.5 ¨³ <= 5
ž¦³Á£šš¸É 3 Áž¨¸É ¥œÂž¨ŠÁ¨È„œo °¥ (¦µ‡µžd—ª´œ™´—Åž – ¦µ‡µžd —ª´œž{‹‹»´ œ)*100 / ¦µ‡µžd—ª´œž{‹‹»´ œ °¥¼n¦³®ªnµŠ -2.5 ™¹ Š 2.5
ž¦³Á£šš¸É 4 ¨ŠÁ¨È „œo°¥ (¦µ‡µžd—ª´œ™´—Åž – ¦µ‡µžd —ª´œž{‹‹»´ œ)*100 / ¦µ‡µžd—ª´œž{‹‹»´ œ > -5 ¨³ <= -2.5
ž¦³Á£šš¸É 5 ¨Š¤µ„ (¦µ‡µžd—ª´œ™´—Åž – ¦µ‡µžd —ª´œž{‹‹»´ œ)*100 / ¦µ‡µžd—ª´œž{‹‹»´ œ <= -5
˜µ¦µŠš¸É 2 ®¨´„Á„–”rš¸ÉčoĜ„µ¦‹ÎµÂœ„ž¦³Á£š®»o œ
˜´ªÂš¸Éčo Ĝ„µ¦ª·‹´¥
‹µ„„µ¦«¹„¬µ™¹ ŠŠµœª· ‹´¥š¸ÉÁ„¸É ¥ª o°ŠÂ¨³¨´„¬–³ o °¤¼¨—´Š˜µ¦µŠš¸É 1 ¡ªn µ¤¸ª·›¸ „µ¦˜nµŠÇ¤µ„¤µ¥š¸É ™¼„
œÎ µ¤µÄoÁ¡ºÉ °ž¦´ž¦»ŠÄ®o„µ¦šÎ µŠµœ °ŠÃ‡¦Š nµ¥ž¦³µšÁš¸¥¤œ´Êœ¤¸ž¦³· š›·£µ¡¤µ„ ¹Êœ ˜n ÁœºÉ°Š‹µ„Šµœª·‹´ ¥š¸É
Á„¸É¥ª o°ŠÁ®¨nµœ´Ê œ ˜´ÊŠ°¥¼nœ“µœ °Š„µ¦œÎµ¦µ‡µ®»oœ °Š˜nµŠž¦³Áš«¤µÄo Ž¹ÉŠÄœŠµœª·‹´¥œ¸Ê ˜o°Š„µ¦Áœo œÁ¡ºÉ°®µ˜´ªÂ
š¸ÉÁ®¤µ³¤„´ ˜¨µ—®»oœÄœž¦³Áš«Åš¥ —´Šœ´Êœ‹¹ŠÅ—ošÎ µ„µ¦š—¨°Š ¹ÊœÁ¡ºÉ° Áž¦¸¥Áš¸ ¥‡ªµ¤Á®¤µ³¤ °Šª·›¸ ˜nµŠÇ
×¥šÎµ„µ¦»n ¤Á¨º°„®»oœ Ĝ˜¨µ—®¨´„š¦´¡¥r®nŠž¦³Áš«Åš¥‹Îµœªœ50 ®»oœ Á¡ºÉ°š¸É‹³®µÂœªšµŠš¸É °—‡¨o°Š„´˜¨µ—
®¨´„š¦´ ¡¥r®nŠž¦³Áš«Åš¥¤µ„š¸É »—Á¡ºÉ°œÎ µÅž¡´•œµ˜n°ÅžÄœ°œµ‡˜ ץĜŠµœª· ‹´¥œ¸Ê ‹³Áž¦¸¥Áš¸ ¥ 2 ¤· ˜·‡º°
‡ªµ¤™¼„˜o °Š ¨³ ‡ªµ¤Á¦Èªš¸Éčo Ĝ„µ¦Á¦¸¥œ¦¼o Á¡ºÉ°š¸É ‹³œÎµÅžÄo Ĝ„µ¦¡¥µ„¦–r˜n°Åž ˜´ªÂ °ŠÃ‡¦Š nµ¥š¸ÉœÎ µ¤µ
š—°¤¸ 3  ‡º° Back Propagation Neural Network, Fuzzy Particle Swarm Optimization with IDDM Neural
Network ¨³ Fuzzy Neural Network ×¥¦µ¥¨³Á°¸¥— °ŠÂ˜n¨³Ã‡¦Š nµ¥ž¦³µšÁš¸¥¤¤¸—´Šœ¸Ê
1. Back Propagation Neural Network ž¦³„°Åž—oª¥®œ¹ÉŠÃ‡¦Š nµ¥ž¦³µšÁš¸¥¤š¸É¤¸ 15  —´Šœ¸Ê
Node 1( ¦µ‡µžd—¥o°œ®¨´Š 1 ª´œ) Node 2(¦µ‡µžd—¥o °œ®¨´Š 2 ª´œ) Node 3(¦µ‡µžd—¥o°œ®¨´Š 3 ª´œ) Node 4(¦µ‡µžd—
¥o°œ®¨´Š 4 ª´œ) Node 5(¦µ‡µžd—¥o °œ®¨´Š 5 ª´œ) Node 6(¦µ‡µžd—¥o°œ®¨´ Š 6 ª´œ) Node 7(¦µ‡µžd —¥o°œ®¨´ Š 7 ª´œ)
Node 8(RSI) Node 9( MACD) Node10(MOMENTUM) Node 11( ADO) Node 12(STOCHASTIC) Node 13(°´˜¦µ
ÁŠ·œž{œŸ¨˜°Âšœ) Node 14(°´˜¦µ®¤»œÁª¸¥œž¦·¤µ–ŽºÊ° µ¥) Node 15(¤¼¨‡nµ®¨´„š¦´¡¥r˜µ¤¦µ‡µ˜¨µ—) ץčo 1
Hidden Layer Ž¹ÉŠž¦³„°Åž—o ª¥ 15 nodes ¨³‡nµLearning Rate Ášnµ„´ 0.5
2. Fuzzy Particle Swarm Optimization with IDDM Neural Network Ĝ‹Îµ¨°Šœ¸ÊčoǦŠ nµ¥ž¦³µš
Áš¸¥¤š´ÊŠ®¤— 6 ǦŠ nµ¥ž¦³µšÁš¸¥¤ ×¥ 5 ǦŠ n µ¥ž¦³µšÁš¸¥¤‹³Äo Áž}œ„µ¦‹ÎµÂœ„˜µ¤ž¦³Á£š®»oœŽ¹ÉŠ
‹ÎµÂœ„×¥¤¸¦µ‡µžd— °Šª´œ™´ —ÅžÁž}œ˜´ ª‹ÎµÂœ„¨³Äo®¨´ „Á„–”r—´Š˜µ¦µŠš¸É 2 ¨³°¸„®œ¹É ŠÃ‡¦Š nµ¥ž¦³µš
Áš¸¥¤‹³Äo Ĝ„µ¦Áž}œ˜´ª‡ª‡» ¤ š´ÊŠ 5 ǦŠ nµ¥ž¦³µšÁš¸¥¤
nªœž¦³„° °ŠÃ‡¦Š nµ¥ž¦³µšÁš¸ ¥¤š´ÊŠ 5 ‹³ž¦³„°—oª¥ input node š´ÊŠ®¤— 15 Á®¤º°œ„´
Back propagation ץčo Hidden Layer 1 Layer Ĝ Hidden Layer ž¦³„°Åž—o ª¥ 15 nodes ×¥¤¸ ‡nµ Threshold
= 0.005, ‡nµ Shake Interval = 500, ‡nµ Shake Area = 0.5, ‡nµ VTR (Value to Reach) = 0.005, ‡nµ maxNFC (Max
Number Function Call) = 500000,‡nµFLC(Falling to Local Control) = 1000, ‡nµ MP (Mutation Point) = 0.02, ‡nµ
Omega, c1 ,c2 = 0.75, 1, 2.5 ˜µ¤¨Îµ—´ (‹µ„„µ¦š—°ÁºÊ°Š˜oœ¡ªnµ particle n ªœÄ®n‹³¤¸‡nµ gbest š¸É—¸„ªn µÁ¤°