整合MRDM、MADM與MODM之互動漸進式粒子最佳化投資組合模型

CCUR > College of Business > Department of Banking & Finance > project > Item 987654321/43997

Please use this identifier to cite or link to this item: https://irlib.pccu.edu.tw/handle/987654321/43997

Title:	整合MRDM、MADM與MODM之互動漸進式粒子最佳化投資組合模型 Interactive Granular-Based Portfolio Optimization Model by Integrating Mrdm, Madm, and Modm
Authors:	沈高毅
Contributors:	推廣部財金系
Keywords:	混合式多準則決策多重法則決策粒子運算約略集合理論互動式多目標最佳化 HMCDM (Hybrid Multiple Criteria Decision-Making) MRDM (Multiple Rule-Based Decision-Making) Granular Computing RST (Rough Set Theory) IMO (Interactive Multiobjective Optimization)
Date:	2018-08~2020-07
Issue Date:	2019-05-03 12:37:58 (UTC+8)
Abstract:	本計畫預計透過創新的混合式架構，提升多準則決策科學在複雜資訊與不確定環境下的決策品質，以及揭露與決策相關、可理解的法則。本計畫偏向基礎導向型研究，試圖結合數種異質性數量方法（包含多準則決策科學、柔性運算與機器學習），透過互動漸進式粒子最佳化來探索決策者偏好，並以建立台股投資組合最佳化作為應用範例。本計畫有別於傳統的多目標最佳化決策模型，主要差別在於假設決策者處於一個具有大量數據結構、複雜度高、充滿不確定的資訊環境；在此前提下，過去複雜的資訊中可能潛藏決策者的偏好組合（結構）；但是基於資訊的複雜與不確定性，決策者無法合理且精準地陳述或指出本身的偏好結構。然而，決策者有機會藉由機器學習與柔性運算的能力，先探勘歷史資料內潛藏的約略知識，找出理性預期空間（包含報酬與風險空間）下的可能組合，逐步提升決策品質。其次，約略知識初始的結構較鬆散，經過漸進式互動，可逐步趨近較精確且近似決策者思維模式的知識粒子結構。為了解釋與驗證此研究構想的可行性，本計畫預計以台灣股市中大型績優權值股為研究標的，協助決策者探索投資選股的約略知識法則，以及探索投資人本身的風險報酬偏好結構。 The project is expected to enhance the decision-making quality of MCDM in complex information and uncertain environments through an innovative and hybrid architecture, as well as to expose the relevant and understandable rules of decision-making. The project is inclined towards basic-oriented research and attempts to explore the preferences of decision makers through interactive progressive optimization by combining several heterogeneous quantitative methods (including multi-criteria decision-making, soft computing and machine learning). The main difference between this project and the traditional multi-objective optimization decision-making model is that the decision makers are assumed to be in an information environment with complicated of data structures, high complexity and uncertainty. Under this presumption, decision-makers may have a hidden preference structure; however, based on the complexity and uncertainty of information, decision makers cannot reasonably and accurately state or indicate their own preferences. However, with the capability of machine learning and soft computing, decision-makers have the opportunity to explore the rough knowledge hidden in historical patterns and find out the possible combinations under the rational expectations space (including return and risk spaces) and gradually improve their decision-making quality. Second, the initial structure of the approximate knowledge is relatively loose. After the interaction, the knowledge structure would be more accurate and close to the thinking mode of the decision maker. In order to explain and verify the feasibility of this proposal, the project is expected to take the blue-chip stocks in the Taiwanese stock market to explore the rough knowledge rules of stock selection and to explore investors' preferences structure.
Appears in Collections:	[Department of Banking & Finance ] project

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	425	View/Open
index.html		0Kb	HTML	274	View/Open

View Licence

Loading...