Full metadata record
DC FieldValueLanguage
dc.contributor.author鄭東雲en_US
dc.contributor.author陳安斌en_US
dc.date.accessioned2014-12-12T02:12:21Z-
dc.date.available2014-12-12T02:12:21Z-
dc.date.issued2003en_US
dc.identifier.urihttp://140.113.39.130/cdrfb3/record/nctu/#GT009134518en_US
dc.identifier.urihttp://hdl.handle.net/11536/58157-
dc.description.abstract在財務預測領域中,人工智慧一向是十分重要的預測工具。但過去人工智慧在財務領域中,一般使用的工具如類神經網路(NN)、基因演算法(GA)等,均需三階段運作:包括訓練,驗證及使用。但這些利用固定歷史資料來訓練以預測未來,均不能即時學習,不符現實需求。本篇文章提出兩個新線上學習演算法,階層預測型分類元系統(SACS)及行為突變預測型分類元系統(ACS3) ,此二演算法均有效協助解財務預測問題。在本文中之實驗證明了SACS及ACS3能在期貨市場的預測上有傑出的表現。在與XCS比較後,驗證了SACS及ACS3有顯著的進步,且能在期貨市場中獲利。 SACS及ACS3是建立在預測型分類元系統(ACS)上,其目的是要分類元系統能適應在波動且複雜的環境,如期貨市場中學習。ACS3是在ACS2中加入action mutation元件,目的是要讓ACS3在不穩定的環境中快速學習。而SACS則在ACS3上再在其輸入中,一改過去二分法而改以使用階層級別分類。這使SACS能有更大的輸入範圍,由其在財務技術指標上,更能突顯其效益。經過本文的驗證,可以確認SACS及ACS3較一般人工智慧方法,更能符合期貨市場預測的需求。而SACS之表現更為突出。zh_TW
dc.description.abstractNowadays, there are many artificial intelligent trading models that comprise of three separate sub-processes: training, validation and application, but these models cannot meet today’s volatile trading environment. Two new online learning algorithms, Scaled Anticipatory Classifier System (SACS) and Anticipatory Classifier System with action mutation (ACS3) are used in futures market trading to satisfy traders’ requirement. This paper aims to proof that SACS and ACS3 provide very good forecast ability in futures market trading performance. Comparing with extended classifier system (XCS), SACS and ACS3 are better suitable to predict in futures market. Finally, the simulation results show that these new algorithm could profit from futures market. ACS3 and SACS are base on Anticipatory Classifier System (ACS). In order to adapt complicate and fluctuant environment, such as future index, ACS3 adds a new component, action mutation, into the algorithm. This leads the population of classifier system learn more quickly and to suit the rapid change environment. SACS has a further improvement by changing the input into scale form; this allows SACS have a more variety input. Particularly for those technical indicators that cannot use binary classification and SACS can solve this problem. After the experiments of this paper, they conclude SACS and ACS3 can reach the requirement of future market. Especially, SACS have a better outstand performance.en_US
dc.language.isozh_TWen_US
dc.subject階層預測型分類元系統zh_TW
dc.subject突變預測型分類元系統zh_TW
dc.subject預測型分類元系統zh_TW
dc.subject分類元系統zh_TW
dc.subjectXCSzh_TW
dc.subject期貨zh_TW
dc.subjectScaled Anticipatory Classifier Systemen_US
dc.subjectAnticipatory Classifier Systemen_US
dc.subjectSACSen_US
dc.subjectACS3en_US
dc.subjectXCSen_US
dc.subjectLearning Classifier Systemen_US
dc.subjectFutures marketen_US
dc.title改良式預測型分類元系統應用在期貨市場之研究zh_TW
dc.titleExploration Enhanced Anticipatory Learning Classifier Systems Used in the Futures Marketen_US
dc.typeThesisen_US
dc.contributor.department資訊管理研究所zh_TW
Appears in Collections:Thesis