社群輿情探勘與主題偵測

本論文整合文字分析與資料探勘技術，分析Facebook社群平台網友們的留言，開發社群輿情分析與主題偵測系統。透過蒐集Facebook社群平台的粉絲專頁和社團中使用者的發文文章與留言內容儲存於資料庫。將文章蒐集大量使用者留言內容進行分析，透過字典比對方式取得留言關鍵字，並運用Entropy過濾掉大部分的General Words，將過濾後的Keywords分別以SVM和Mining Association預測Topic Keywords，比較兩者的準確度，並以準確度高的方法套用在預測Topic Keywords。統計Topic Keywords各個時間點上的聲量並以時間軸方式呈現。最後在由情緒字典，正負面關鍵字以分數加總方式推測留言情緒是正面或負面。

關鍵字

資料探勘；粉絲專頁；主題關鍵字；情緒分析

並列摘要

This thesis integrates text analysis and data exploration technology, analyzes the messages of Facebook community platform users, and develops a community public opinion analysis and topic detection system. Through the collection of the fan page of the Facebook social platform and the posts and comments of community users, the content is stored in the database. The article collects user comments and analyzes, obtains the keywords of the comments through dictionary comparison, and uses Entropy to filter out most of the General Words. Use SVM and Mining Association to predict Topic Keywords.Compare the accuracy and apply a highly accurate method to predict Topic Keywords.Count the volume of Topic Keywords at each time point and present it as an information graph.Finally, from the sentiment dictionary, the positive and negative keywords are summed up to infer whether the message sentiment is positive or negative.