在這個資訊爆炸的這個時代，每個人都能輕易的在網路上取得或發佈資訊，能夠快速的篩選出正確且有用的訊息是ㄧ件很有意義及重要的事情，而意見領袖的意見通常能代表大部分的人，其意見可信度高且會有大量追隨者被影響，因此會希望能透過意見領袖來獲取或者是發佈資訊。 本研究提出一個雙分群意見領袖探勘法, 稱為TCMOLM (Two-Clustering Method for Opinion Leader Mining, TCMOLM)，同時結合了圖論與特質方法來尋找意見領袖，並將意見分析方法運用在最後得出意見領袖，去判斷意見領袖的傾向，可使得意見領袖在市場行銷方面的運用上能更靈活。 根據汽車領域方面的實驗結果，證實TCMOLM能有效降低影響力重覆的問題，其效能高於單純使用特質方法來尋找意見領袖。 In the modern age where information explodes every minute, everyone has an easy access to all sorts of information, and they can easy publish their idea on the in-ternet. To quickly filter out the right message is a useful and very meaningful thing, we can obtain or release information through opinion leaders, because opinion leaders can affect a lot of people, and their idea is high reliability and represent the majority of the people. This research presents a method (Two-Clustering Method for Opinion Leader Mining, TCMOLM) combined with graph theory and methods which use characteris-tics to find opinion leaders and coordinate semantic analysis methods to detect opin-ions tendency of opinion leader. We can use opinion tendency to let marketing more flexible. According to the results of experiment on the automotive sector aspect, which is confirmed by TCMOLM, can effectively reduce the impact of repeated problems. The performance of our research is higher than the method used characteristics only to find opinion leaders.