English  |  正體中文  |  简体中文  |  Items with full text/Total items : 52047/87178 (60%)
Visitors : 8683456      Online Users : 84
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library & TKU 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://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/105121

    Title: 群眾智慧的代理人基建模 : 意見形成模型的應用及比較
    Other Titles: Agent-based modeling of the wisdom of crowds : application and comparison of opinion formation models
    Authors: 古承翰;Ku, Cheng-Han
    Contributors: 淡江大學產業經濟學系碩士班
    池秉聰;Chie, Bin-Tzong
    Keywords: 代理人基建模;群眾智慧;意見形成;種族分離模型;傳染病模型;投票模型;選民模型;Agent-based modeling;Wisdom of Crowds;Opinion formation;Segregation model;Virus model;Voting model;Voter model
    Date: 2015
    Issue Date: 2016-01-22 14:48:16 (UTC+8)
    Abstract: 本文目的在研究零碎的個人資訊如何透過互動的機制,達成意見共識的形成。我們使用代理人基模擬方法分析比較五種常見的意見形成模型,包括易辛模型(Wu and Szeto的版本)、種族分離模型、傳染病模型、投票模型,以及選民模型。由於不同模型之間的參數設計及互動方式各異,我們先針對五種模型內的特定參數組合對達到共識或穩定狀態所需時間進行分析。接下來,對模型共通參數進行跨模型的比較,這些比較包括代理人人數以及意見數對各模型達到共識或穩定狀態所需時間。為了體現索羅維基 (James Surowiecki) 的「群眾智慧」概念,我們選定一個較短的時間,目的在於觀察各模型尚未達到共識時的狀態,與初始設定狀態的差異。探討不同的意見形成模型中,在時間壓力下意見形成是否具有「群眾的智慧」,亦或只是單純的從眾行為。
      本研究得到以下結果,就達成穩定時間而言:(1) 在Wu and Szeto模型中,個人的社會互動性越強,平均達成共識時間越長。當代理人適度與同質性的他人互動時,人口達到共識的速度非常快。而與同質性他人互動太高或太低,達到共識的時間反而更長;選民模型中,在局部小區域互動之下,當代理人意見更新的比例越高,平均達成共識時間越快;投票模型中,也是在局部小區域互動,當純粹考慮週遭訊息時,最後共識將偏向優勢意見。考慮代理人同情弱勢意見時,結果將相當不穩定。考慮游移不定的代理人時,則更容易偏向優勢意見。同時考慮上述兩種代理人時,社會呈現兩極化的狀態;種族隔離模型中,當人數越多,到達穩定狀態的時間也將越長。當與同質性越高的代理人互動時,到達穩定狀態的時間也將越長;傳染病模型中,我們發現人數越多並不會導致傳染病模型的平均達成共識時間顯著的增加,得到較快達成共識時間的反而是介於中等密集程度的50%。而當代理人宣傳自己意見的機率愈高,平均達成共識時間就越快。(2)就代理人人數而言:除了Wu and Szeto模型外,其他模型代理人人數對平均達到共識時間或達到穩定狀態時間的影響皆為正,影響的幅度為選民模型 > 傳染病模型 > 種族分離模型。而代理人人數對Wu and Szeto模型達到共識時間的影響呈現倒U型。(3)就意見數而言:在Wu and Szeto模型中,三意見的平均達到共識時間通常比兩意見在同樣的模型來得更快。但在其它模型中兩意見的平均達到共識或達到穩定狀態時間會比三意見來得快。(4)群眾智慧:選民模型和傳染病模型相較於其他三個模型,有類似於「群眾智慧」的形成,我們認為這是由於選民模型和傳染病模型中的代理人與其他代理人互動的隨機性,導致這兩種模型的代理人在改變自己的意見時,較不受到其它代理人的影響,這類似於形成群眾智慧四個條件之中的「獨立」,因此各意見的誤差可能會互相抵銷,導致經過一段時間後,三個意見的平均比例與初始設定值相同。
    The purpose of this thesis is to study how the consensus is emerging forms from individual fragmented information through interaction. We use an agent-based simulation to analyze and compare the performances of five opinion formation models, including Ising model (version of Wu and Szeto), segregation model, virus model, voting model, as well as voter model. Since parameters design and interaction mechanism between these models are different, we first analysis the required time for reach consensus or steady state under specific parameter combinations. Next, we compare common parameters across models, these comparisons include the required time for reach consensus or steady state of the number of agents and the number of opinions for each model. In order to emerge the concept of "wisdom of crowds" by James Surowiecki, we designed a “time constraint” to observe the different performances of each model. To analysis of different opinion formation models, we examine the opinions formed under time pressure to observe whether they have met the criterion of "wisdom of crowds", or just simply "herding" behavior emerged.
      We summarize our fruitful results as follows: (1) For the time to steady-state aspect: In Wu and Szeto model, the higher social interaction, the longer the average time to reach a consensus. When the homogeneity of interaction is moderate, the population can reach a consensus very quickly. When the homogeneity is small or large, the population significantly need longer time to reach consensus; In voter model, under the local interaction, the higher percentage of information update, the faster the average time to reach a consensus; In voting model, when local interaction is applied, the trend to an advantage opinion situation is evident. When considering the sympathy effect, the system will show an unstable state. When considering the median voter effect, the system will convergence to an advantage opinion situation is more obvious. When considering both effects, the system will show a polarization state; In segregation model, the larger the number of agents, the longer time to reach steady state will be need. When agents want to interaction with homogeneous neighbors, the time to reach steady state will be longer; In virus model, we found that the number of agents does not cause a longer time to reach a consensus state. To get a faster convergence time is to increase the percentage of infectiousness (idea propaganda) (2) Number of agents: In addition to the Wu and Szeto model, other models affect the number of agents on the average time to reach a consensus or the time to reach steady state are all positive, the magnitude of the impact is voter model > virus model > segregation model. (3) Number of opinions: In Wu and Szeto model, the average time to reach a consensus of three-opinion model is generally more quickly than the one of two-opinion model. But in other models, the average time to reach a consensus of two-opinion model is faster than three-opinion one. (4)Wisdom of crowds: Among the five models, voter model and virus model have the results similar to the formation of "Wisdom of Crowds," we believe that it is due to the two models that agents randomly interact with each other. Therefore, in these two models, when agents updating their opinions, there are less influenced by others, which is similar to one of the four conditions of the formation of the Wisdom of Crowds, so the error of each opinion could offset each other, resulting after a period of time, the average ratio of the three opinions and the initial setting almost identical.
    Appears in Collections:[產業經濟學系暨研究所] 學位論文

    Files in This Item:

    File Description SizeFormat

    All items in 機構典藏 are protected by copyright, with all rights reserved.

    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library & TKU Library IR teams. Copyright ©   - Feedback