以線上評論的意見分析建立導航行動軟體的體驗品質評估模型

GPS接收器已成為行動裝置的基本內建配備，在過去的幾年裡，許多行動軟體開發者以這項配備為基礎，推出了他們自己的產品以提供導航服務。如今，導航行動軟體已成為許多行動用戶生活中不可或缺的工具。透過導航行動軟體，使用者可以很快地找到路徑以前往其想要到達的地方，途程中並且可以享受推薦的服務及拜訪有趣的景點。導航類行動軟體的蓬勃發展雖然帶給了消費者多元化的選擇，但也使開發者之間的競爭不斷加劇。為了強化自身產品及服務的競爭力，開發者因此需要花費許多精力在提高使用者的體驗品質，一方面用以增加使用者的忠誠度，另一方面吸引更多的使用者使用。線上評論是現今許多使用者在決定是否購買特定軟體時會參考的資訊，這些評論來自使用者體驗過特定軟體後的心得，具有相當程度的影響力，也是開發者改善自身產品的參考依據。線上評論數量龐大且非結構化，因此透過閱讀以萃取重要資訊相當困難。本研究對導航類行動軟體收集Apple's App Store Taiwan的線上評論，以意見分析與深度學習法從評論中推導出最顯著影響使用者體驗滿意度的30個特徵品質，據以建立導航類行動軟體的體驗品質評估模型。本研究並以adjusted R2 (調整後決定係數) 、RMSE(均方根誤差)及MAPE(平均絕對百分比誤差)對該模型的自變數的解釋力及預測能力進行評估。評估結果顯示adjusted R2、RMSE及MAPE各為0.9739、0.29及5.647%，此結果優於早先鄭育儒(2015)等所使用的逐步線性回歸分析。本研究所建立的模型能應用以提供消費者從中了解特定導航軟體的品質性能以進行下載決策，也能使開發者確認其產品及服務的改進方向。本論文最後將對研究成果進行討論，並提出建議。

關鍵字

導航類行動軟體；線上評論；體驗品質；深度學習；意見分析

並列摘要

GPS receivers have become the basic built-in feature of smartphones. With the feature, numerous app developers launched their apps over the past years to provide navigation services. Consequently, navigation apps have been mobile users’ indispensable tool for finding their ways to destinations, enjoying recommended services or visiting interesting spots. The rapid growth of navigation apps has been diversifying consumer choices, as a result, the app developers’ competition has been intensified. Developers thus spend considerable efforts to enhance users’ Quality of Experience (QoE) so as to maintain users’ loyalty and to attract more consumers. Writing by experienced users, online reviews are the critical information resources that influence a user’s app purchase decision. Moreover, the online reviews recommend the improvement suggestions to the developers. The online reviews are unstructured, uneven in quality and massive in quantity, therefore, it is difficult to extract valuable information from them. We collected online reviews of Navigation app form Apple's App Store Taiwan, used opinion analysis and Deep Learning method to derive 30 most influential quality features. Accordingly, this study developed a quality of experience assessment and prediction model for Navigation Application. We tested the explanatory and predictive powers of the model using adjusted R-squared (R2, the coefficient of determination), Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE), which were 0.9739 and 0.29 and 5.647% in order. Furthermore, a comparison between the performance of Deep Learning and the Stepwise Linear Regression used by Zheng (2015) shows that the former outperformed the latter. With the QoE model, practical applications demonstrated how users and developers can benefit from it to respectively make download decisions and identify apps’ improvement directions.

並列關鍵字

Navigation apps ； online review ； quality of experience ； Deep Learning ； opinion analysis

參考文獻

1. 李易霖. (2014). 萃取線上評論品質特徵以發展行動軟體搜尋法. (碩士), 國立屏東科技大學. 屏東縣.

Google Scholar

2. 江明晏. (2015). 蘋果App Store下載數突破1000億次. Retrieved November 25, 2015, from https://tw.news.yahoo.com/%E8%98%8B%E6%9E%9Capp-store%E4%B8%8B%E8%BC%89%E6%95%B8-%E7%AA%81%E7%A0%B41000%E5%84%84%E6%AC%A1-032935792--finance.html

Google Scholar

3. 吳明隆. (2003). SPSS 統計應用學習實務—問卷分析與應用統計, 台北: 知城.

Google Scholar

4. 呂怡萱. (2010). 運用 GPS 分析市區公車站位使用頻率及因素研究.

Google Scholar

5. 周欣穎. (2012). 台灣旅遊消費結構與滿意度之研究. (碩士), 真理大學, 新北市.

Google Scholar

主題瀏覽