三元決策理論應用於國道計程收費議題之情感分析研究

Internet Public Opinion Sentiment Analysis on Topic of Taiwan Freeway’s Distance-based Toll Collection Using Three-way Decisions Theory

蔡易辰(Yi-Chen Tsai)

指導教授 ： 陶治中

淡江大學/商管學院/運輸管理學系碩士班/碩士(2016年)

https://doi.org/10.6846/TKU.2016.00892

摘要

交通運輸一直是人類生活需求當中不可或缺之環節，各種交通政策執行績效的成敗與否，使用者網路意見蒐集與即時化解民怨的變化往往影響整體政策的發展大局。實施國道高速公路計程電子收費政策之後， ETC整體用路人滿意度在2015年雖達七成，政策執行過程仍產生諸多爭議及問題，短期間網路上常湧現大量的評論，使得部分民眾消極使用；同時新興媒體的出現，也使得輿情更加多元更具時效性，政府應當正視並避免危機擴大，對ETC使用者之想法、意見與經驗等資訊進行有效的蒐集與分析，將有利於營運管理者能即時掌握重要議題，進而擬訂相對應之策略並有效化解無法預期的危機。

有鑒於此，本研究係使用爬蟲系統蒐集輿情文本，以網路評論意見為資料基礎，進行意見挖掘並建構三元決策情感分析模式。本研究運用三元決策理論於情感分析，輿情文本情感傾向將可具有正面情感、中立情感與負面情感，以較精準的分類方法，輿情文本依其內容計算情感值，並根據情感區間門檻值劃分正面、負面與中立區域。輿情文本情感值計算以正、負面情感詞庫為情感值計算之依據，利用詞頻演算法計算輿情文本的情感值，透過三元決策情感分析模型進行情感分析，最終計算出三元決策情感分析模型之通用模型與情感區間門檻值。在與遠通電收公司達成保密協定下取得客服語音資料，進行ETC使用者對ETC服務項目抱怨程度之分析。

本研究利用三元決策情感分析模型對網路評論文本進行分析，而客服語音資料僅藉由計算情感值以探討民眾抱怨之情感程度。研究結果顯示，ETC網路輿情八大構面與主題之情感分析部分，正面情感傾向之主題以交管措施為主，中立情感傾向之主題以申裝異動/車輛過戶與供裝，負面情感傾向之主題絕大多數為收費員議題；ETC客服語音六大構面與主題之負面情感程度分析部分，繳費服務構面之繳費逾期主題跟平信掛號與通知主題，以及申辦服務構面之查詢管道與流程主題，其為遠通電收公司較需注意與改善之服務項目。ETC網路輿情與客服語音之情感認知迴歸分析，網路輿情之檢定結果係以政策構面為特別顯著構面，而客服語音之檢定結果顯示繳費服務構面為特別顯著構面，其為半年期間內特別會影響ETC整體情感趨勢變化之構面。建議後續情感認知迴歸式之更新計算宜以年度為時間單位，將使整體分析結果更為完整。

關鍵字

三元決策理論 ； 情感分析 ； 情感認知迴歸分析 ； 國道計程電子收費

並列摘要

Transportation has always played as a significant role of modern human life, however, the acceptance of public transportation policy is evaluated by user satisfaction. User’s opinion, satisfaction and problem-solving efficiency should be taken into consideration by policies making department. In 2015, the user satisfaction of Freeway's Distance-based Toll Collection( ETC) of Taiwan was higher than 70%, however, there are bias and dispute existed during the implementation of policy. Users easily are affected by negative comments through internet to boycott using ETC, meanwhile various electronic media allow public opinion spreading quickly. Minisrty of Communication and Communications(MOTC) should avoid the gap between polices and users expectation by collecting and analyzing opinion into reference for policy making department and reduce the unpredictable problems caused by public opinions.

　　This study used text data mining for collecting data and analyzing the public opinions, comments and sentiment tendency and providing suggestions to policy making and operation management teams. The data collected from opinion text were by Crawler system. The opining text based on user’s comments on network is for opinion mining and sentiment analysis. This research used three-way decisions sentiment analysis model by using three-way decision theory to analysis sentiment analysis, opinion text emotional tendencies which include : positive, boundary and negative, which categorized by calculating the sentiment value, divided into 3 sentiment zones ; positive, negative and neutral zone . The calculation of sentiment values were based on positive and negative sentiment dictionary and words frequency algorithm opinion text. By using three-way decision model for sentiment analysis and calculated universal of three-way decisions sentiment analysis model, and “threshold of sentiment zone”. Finally, this study obtained privacy agreement and authorized using customer’s voice data from FETC to analysis ETC users complaint problems.

　　The results show that , regarding ”eight of ETC Internet sentiment dimensions” and “sentiment analysis”, the topic of positive sentiment is”traffic control” , the topic of neutral sentiment is “install and modify/transfer of vehicle ownership and Installation”, the topic of negative sentiment is “toll collector issue”. The six voice dimensions of ETC customer service center and topic of negative sentiment ; “late payment”, “ordinary mail” , ”notification”, “application service” and “discover the pipeline and process” need to be improved. According to results of sentiment cognition regression analysis, ETC Internet opinion, voice public opinion and customer voice , the “policy” is significant dimension, the test result of “customer voice” shows “payment service” is significant dimension, which has significant effect on ETC overall sentiment dimensions in following 6 months. This study suggests that using annual sentiment cognition regression updated calculations will obtain more detailed results.

並列關鍵字

Three-way decisions ； Sentiment Analysis ； Sentiment Cognition Regression Analysis ； Distance-based Electronic Toll Collection

參考文獻

10.林千斐(2009)，使用R軟體計算Simon二階段第二期癌症臨床試驗設計的樣本數，中原大學應用數學系碩士學位論文。

16.陳亭愷(2015)，國道計程電子收費實施後之網路輿情文本情感分析研究，淡江大學運輸管理學系運輸科學碩士班碩士論文。

17.許中川、陳景揆(2001)，「探勘中文新聞文件」，資訊管理學報，第7卷第2期，頁103-122。

19.莊依珊(2015)，滿意度指數為基礎之國道計程電子收費服務預警機制研究，淡江大學運輸管理學系運輸科學碩士班碩士論文。

26.簡之文(2012)，部落格文章情感分析之研究，淡江大學資訊管理學系碩士班碩士論文。