English  |  正體中文  |  简体中文  |  Items with full text/Total items : 66984/66984 (100%)
Visitors : 23058028      Online Users : 218
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Tips:
  • please add "double quotation mark" for query phrases to get precise results
  • please goto advance search for comprehansive author search
  • Adv. Search
    HomeLoginUploadHelpAboutAdminister Goto mobile version


    Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/81254


    Title: 卷積模糊神經系統於財金市場數據之研究;CNN Neural Fuzzy Time Series Forecasting in Financial Markets
    Authors: 許敦盛;Hsu, Dun-Sheng
    Contributors: 資訊管理學系
    Keywords: 多目標時間序列預測;卷積神經網路;模糊神經系統;球型複數模糊集;複合式最佳化演算法;multi-targets time series prediction;convolution neural networks;neural fuzzy system;sphere complex fuzzy sets;hybrid optimization algorithm
    Date: 2019-07-16
    Issue Date: 2019-09-03 15:40:37 (UTC+8)
    Publisher: 國立中央大學
    Abstract: 財金市場的時間序列預測是個具挑戰性的議題。在過去的研究中,多數方法只針對單一目標進行預測。因此,本研究提出多目標預測方法,結合卷積神經網路(Convolutional neural networks, CNN)與球型複數模糊神經系統(Sphere complex neural fuzzy system, SCNFS)。CNN能從模型輸入中,擷取出有用的資訊,藉此提升模型的預測表現;球型複數模糊集(Sphere complex fuzzy sets, SCFSs)能產生多個複數型態的歸屬程度,此特性使本研究所提之模型擁有進行多目標預測的能力。在前處理階段,使用多目標特徵選取,從輸入資料中挑選有影響力之特徵作為模型輸入。參數訓練方面,使用高斯分布型鯨群最佳化演算法(Gaussian distribution based whale optimization algorithm, GD-WOA)及遞迴最小平方估計法(Recursive least squares estimator, RLSE)之複合式最佳化演算法,透過分而治之的方式,GD-WOA針對CNN卷積核參數及球型複數模糊集參數進行訓練;RLSE針對SCNFS之後鑑部層參數進行訓練。為驗證本研究所提之方法,三個實驗以股價指數數據作為資料集,進行不同目標數的時間序列預測。實驗結果與過往文獻相比,有較佳的預測結果,顯示本研究所提之方法,在時間序列預測上有良好效能。;Time series prediction of financial markets is a challenging issue. In literature, most prediction methods only predict for a single target at a time. In this study, a novel method for multi-target prediction is proposed, integrating a convolutional neural networks (CNN) and a sphere complex neural fuzzy system (SCNFS), called CNN-SCNFS. The membership degree of sphere complex fuzzy sets (SCFSs) can accommodate more membership information than the conventional fuzzy sets, enabling the SCNFS prediction model to achieve multi-output property. In the pre-processing stage, cross-target feature selection is used. However, for the SCNFS, the increasing number of input features usually result in the addition of computational resources on training. To receive the more number of informative features, a CNN architecture is placed in front of the SCNFS. Due to capability of capturing sudden changes, CNN is a suitable model for the chaotic situation. In terms of optimization algorithm, the concept of divide-and-conquer is applied. Gaussian distribution based whale optimization algorithm (GD-WOA) tunes the kernel maps of CNN and the parameters of SCFSs; recursive least squares estimator (RLSE) tunes Takagi-Sugeno layer of SCNFS. Great performance is shown by the GD-WOA-RLSE hybrid optimization algorithm. The trading information of different stock index is used as the data sets, to perform the experiment of multi-target prediction. The experimental results indicate that the proposed method is comparable with the other compared methods on the time series prediction.
    Appears in Collections:[資訊管理研究所] 博碩士論文

    Files in This Item:

    File Description SizeFormat
    index.html0KbHTML17View/Open


    All items in NCUIR are protected by copyright, with all rights reserved.

    社群 sharing

    ::: Copyright National Central University. | 國立中央大學圖書館版權所有 | 收藏本站 | 設為首頁 | 最佳瀏覽畫面: 1024*768 | 建站日期:8-24-2009 :::
    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback  - 隱私權政策聲明