- 學位論文
社群媒體大數據分析應用於臺鐵安全績效與網路輿情對照之研究
Mapping TRA's Safety Performance with Internet Public Opinion by Social Media Analytics
摘要
近年來隨著社群媒的蓬勃發展，民眾經常使用社群媒體網路來表達自己的觀點，社群媒體儼然成為民眾發表意見、抒發情緒、彰顯自身觀點與立場的重要平臺。再加上大數據分析的流行，衍生出社群媒體挖掘技術，藉由大數據挖掘，分析民眾觀點與情感傾向，可更快速理解民眾需求。 本研究欲建立一網路輿情情感分析模式，透過蒐集社群媒體言論，運用文本分類與資料探勘技術，並將評論主題分為「車站設施」、「員工權益」、「餐飲服務」、「票證系統」、「列車運轉」五大類相關文本，分析民眾關注之臺鐵服務相關議題與輿情情感趨勢，再經由情感分析結果計算社群媒體留言之情感值，最後與臺鐵提供的安全績效(亦即事件發生資料)進行對照，以驗證本研究結果之正確性。 經由實證分析結果得知，五類評論中僅有列車運轉類別在關聯模型中最具有顯著性。平均情感分數最低的評論類別為票證系統，餐飲服務雖表現最好，但平均情感分數仍為負數，顯示臺鐵的服務仍有改善空間。 本研究借助視覺化技術，將安全績效與網路輿情兩者合併成對照圖，以時間軸檢視安全績效與網路輿情資料，探討相同時間點兩者資料的相關聯，亦即當事件或事故發生時，是否會對網路輿情的情感趨勢產生變化，以供相關單位參考之用。
並列摘要
In recent years, people have used social media networks frequently to express their opinions because of vigorous development of social media. Therefore, social media mining with big data analytics can be applied to overview public opinions and sentiment tendencies for the purpose of intelligent decision making. This study aims at establishing an internet public opinion analysis model with text mining technologies. Firstly, social media comments on popular websites are collected. Text classification approaches are then used to divide TRA‘s (Taiwan Railway Administration) service related comments into "station facilities", "employee rights", "catering services", "tickets system" and "train operation" types of topics. And then emotional values of daily comments on these five topics are calculated based on sentiment analysis. Finally, safety performance data provided by TRA (I.e. incident data) are chosen for mapping with internet public opinion results. Empirical results showed that only "train operation" is the most significant category in the correlation with safety performance. And the lowest average sentiment score is "tickets system" category. Although "catering services" category shows positive comments, its average sentiment score is negative. It indicates that TRA’s services need more active improvements. A comparison chart can be visualized to map TRA’s safety performance with internet public opinions. It is proven to be helpful for TRA’s decision makers to monitor public opinion changes by using social media mining when accidents or incidents happen.