政大機構典藏-National Chengchi University Institutional Repository(NCCUR):Item 140.119/53431
English  |  正體中文  |  简体中文  |  Post-Print筆數 : 27 |  全文笔数/总笔数 : 109948/140897 (78%)
造访人次 : 46072187      在线人数 : 699
RC Version 6.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
搜寻范围 查询小技巧：
  • 您可在西文检索词汇前后加上"双引号"，以获取较精准的检索结果
  • 若欲以作者姓名搜寻，建议至进阶搜寻限定作者字段，可获得较完整数据
    •  进阶搜寻


    请使用永久网址来引用或连结此文件: https://nccur.lib.nccu.edu.tw/handle/140.119/53431


    题名: 以資料為中心之類神經網路方法論與離群值
    其它题名: Neural-Networks-Related Data-Based Approach and Outliers
    作者: 蔡瑞煌
    贡献者: 國立政治大學資訊管理學系
    行政院國家科學委員會
    关键词: 資料;類神經網路;方法論;離群值
    rare events;stock market;single-hidden layer feed-forward neural networks;the resistant learning context
    日期: 2011
    上传时间: 2012-08-30 15:49:06 (UTC+8)
    摘要: 股市生產著大量的日數據，而其輸入輸出關係通常是非線性的、且未知的。但在每日的 股市指數，也有罕見事件（即遠離主體數據點之離群值）。當我們希望通過直接使用收 集到的股市指數，並且採用類神經網路，來進行系統建模、優化和控制時，要從每日的 股市指數裡，系統性地確認罕見事件仍是一問題；而要有效地處置罕見事件則是另外一 問題。 (Tsaih and Cheng, 2009)採用單一隱藏層倒傳遞神經網路(SLFN)為建模工具來說明 以下之穩健學習情境：  在學習過程中，SLFN 的隱藏節點數量是可調整的  在學習過程中，離群值是遠離所得之 SLFN 函數的數據點  潛在的離群值不該在學習的過程中顯著地影響 SLFN 的隱藏節點數量 (Tsaih and Cheng 2009)的穩健學習過程可以把數據分成主體數據點群和潛在的離群值 群。 目前沒有任何研究於穩健學習情境下之有效的以資料為中心之類神經網路方法論 的事實激發了本研究。具體來說，本研究的目標是： (1)設立股市實驗，以探索在穩健學習情境下之有效的神經網絡相關的數據為基礎的方 法。更具體地說，在數據預處理階段，採用(Tsaih and Cheng 2009)的穩健學習過程 來把訓練數據分成主體數據點群和潛在的罕見事件群。在建模階段，採用(Rumelhart, Hinton and Williams 1986)的學習算法調整SLFN 模型。在預測階段，應用取得之 SLFN 模型於預測數據，以獲得相應的預測。 (2)探索精選的罕見事件的相關議題。也就是說，我們要探索以下一些問題：精選的罕 見事件是否是真實的罕見事件？能否有任何進一步的分析這些精選的罕見事件，來 幫助理解如何檢測他們以及他們對股市的影響？ (3)驗證在數據預處理階段，採用(Tsaih and Cheng 2009)的穩健學習過程的必要性。
    In modern stock markets whose input-output relationship is normally non-linear and unknown, vast amounts of data are being produced each day. But there are rare events (outliers that are far from the bulk of observations) in the daily index of stock market. When we adopt the Neural-Networks-related (NN-related) data-based approach to perform the system optimization, control and modeling via directly using the input-output data collected from the stock market, to systematically identify the rare event from the daily index of stock market is a study issue and to effectively cope with the outlier is another study issue. With single-hidden layer feed-forward neural networks (SLFN) as the modeling tool, Tsaih and Cheng (2009) describe the resistant learning context as follows:  the number of adopted hidden nodes of SLFN is adaptable within the learning process,  outliers are the observations far away from the (fitting) function defined by the obtained SLFN, and  potential outliers should not impact significantly the number of adopted hidden nodes within the learning process. Furthermore, their resistant learning procedure can separate the data into two subsets, the bulk and the potential outlier. The fact that there is no study of data-based approach in the resistant learning context motivates this study. Specifically, the objectives of this study are: (1) Set up a stock-market experiment to explore an effective NN-related data-based approach for coping with rare events in the resistant learning context. More specifically, in the data-preprocessing stage, the resistant learning procedure of (Tsaih & Cheng, 2009) is adopted to separate the training data into two subsets, the bulk and the (potential) rare events. In the modeling stage, the bulk data is used to tune the SLFN model via the back propagation learning algorithm (Rumelhart, Hinton and Williams 1986). In the forecasting stage, the obtained SLFN model is applied to the forecasting data to get corresponding predictions, based upon which investment decisions are made. (2) Explore the reality of “picked” rare events in the stock market. That is, we want to explore some of the following questions: Are some of the “picked” rare events real ones of stock market? Any further analysis of them can help understand the way of detecting them and the way they affect the stock market? (3) Justify the necessity of data preprocessing mechanism for coping with rare events.
    關聯: 應用研究
    學術補助
    研究期間:10008~ 10107
    研究經費:348仟元
    数据类型: report
    显示于类别:[資訊管理學系] 國科會研究計畫

    文件中的档案:

    档案 描述 大小格式浏览次数
    100-2410-H-004-013.pdf1158KbAdobe PDF2733检视/开启


    在政大典藏中所有的数据项都受到原著作权保护.


    社群 sharing

    著作權政策宣告 Copyright Announcement
    1.本網站之數位內容為國立政治大學所收錄之機構典藏，無償提供學術研究與公眾教育等公益性使用，惟仍請適度，合理使用本網站之內容，以尊重著作權人之權益。商業上之利用，則請先取得著作權人之授權。
    The digital content of this website is part of National Chengchi University Institutional Repository. It provides free access to academic research and public education for non-commercial use. Please utilize it in a proper and reasonable manner and respect the rights of copyright owners. For commercial use, please obtain authorization from the copyright owner in advance.

    2.本網站之製作，已盡力防止侵害著作權人之權益，如仍發現本網站之數位內容有侵害著作權人權益情事者，請權利人通知本網站維護人員(nccur@nccu.edu.tw)，維護人員將立即採取移除該數位著作等補救措施。
    NCCU Institutional Repository is made to protect the interests of copyright owners. If you believe that any material on the website infringes copyright, please contact our staff(nccur@nccu.edu.tw). We will remove the work from the repository and investigate your claim.
    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - 回馈