資訊負荷已成為現今資訊時代的一大問題，在資訊量成長快速的狀況下，資訊過 濾的方式也叫無效率。另一方面，社會網絡中的使用者越來越傾向將網路上的內容排 序。本研究主要目的在探討是否可衡量社會網絡中的信任值，以及透過分群的方式是 否能有效協助分析。此外，本研究也提出信任值計算模式，此模式以社會網絡以及線 上評分系統為基礎。本研究透過社會距離的概念，輔以分群法為基礎協助區隔社會網 絡中的個體。並以分群後的個體間距離為信任值推算基礎。 研究結果顯示透過分群法應用在社會網絡中，是能夠推算出信任值。本研究所 提出的模式結合了不同變數如時間變數與評分值的概念，在計算信任值上以多維度的 角度來思考。研究結果也顯示較高的評分結合較短的社會距離夠產生較高的信任值， 較低的評分結合較長的社會距離則產生較低的信任值。這也證明了社會網絡中社會距 離的影響性。縱言之，本研究主要提供了多維度的信任值推算模式，主要考量變數包 含社會距離、權重、時間以及內容評分值，此外，不同層級的個體關係也有不同的重 要性。本研究期望對於社會網絡中信任的概念文獻提供初步的基礎，並且也提供在資 訊爆炸時代下資訊過濾方式的一個新思維。 Information overload is an increasing problem, and as information available continues to grow in volume, current filtering techniques are proving inefficient. Social network users and people in general, tend to prioritize recommendations coming from people they are acquainted to. The purpose of this study was to investigate if it was possible to measure trust within individuals in a social network, as well as find out if data clustering methods could help to achieve said goal. Another aim was to develop a trust model that would estimate a trust value for content creators on an online rating system with social network capabilities. This research introduces the concept of social distance, which is drawn from clustering methods applied to the social network user base; and incorporates said distance in the estimation of trust, as well as user generated ratings. The trust value estimated will serve as a metric for filtering and sorting content of any kind based on the trustworthiness of the creator. The results of the study revealed that it is possible to provide an estimate measure of trust within individuals in a social network and that clustering methods were of significant help into said evaluation as well as the integration of other variables affecting the building of trust. It was found that the model proposed by this study was able to integrate various variables and provide a more complete and integrated, multidimensional value to an estimated trust. Results also showed, that higher rating scores combined with shorter social distances provide satisfactory trust values, while the opposite happened for subjects presenting lower rating scores in combination with longer distances. The principal conclusion was that our model provides a multidimensional estimated value for trust on content from the Internet, that integrates some of the variables necessary for the building of trust in online setting, as are: social distance, weight of relationship, time, and ratings from an online rating system; as well trust levels between individuals within a social network. This study contributes to the current literature on trust estimation and social networks role in such endeavors. This will provide also an alternative for current information overload issues as well.