建構個人化服務人員與小費估計之推薦機制

A Hybrid Approach for Personalized Tour Guide Service and Tip Estimation

容健芳(Chien-Fang Jung)

指導教授 ： 張瑋倫

淡江大學/商管學院/企業管理學系碩士班/碩士(2014年)

https://doi.org/10.6846/TKU.2014.01033

摘要

服務業的成長在經濟發展中顯著而易見，而服務中人員所造成的服務異質性仍是服務提供中維持服務品質的一項難解的難題之一。過去許多研究探討並試圖解決因雙方（服務人員與顧客）的個人因素所產生的品質不穩定，包括了服務人員的工作適配度、員工教育訓練以及工作激勵等角度進行討論；也討論了顧客在服務流程中扮演的角色以及服務預期與服務體驗差距產生的服務認知價值，但卻鮮少討論到服務傳遞者與服務參與者彼此間的互動關係所導致的服務品質差異。
本研究將研究範圍聚焦在觀光導覽服務上，建立一套新的服務流程，將服務精神回歸到以人為本。利用自我組織映射分群法以及協同式過濾推薦法，建立一套全新的機制，將服務人員的選擇及小費金額的給予兩種服務程序，緊扣於顧客認知價值下的選擇。透過自我組織映射法中非監督式的學習網絡，將顧客資料進行分群，來把推薦基準拉近至特質相似的顧客群。再透過協同式過濾推薦法，利用過去顧客的偏好進行新顧客的偏好預測，來推薦合適的服務人員。最後利用自我組織映射分群法，找出被同一服務人員服務過的顧客進行分群，計算出建議的小費金額。在此機制的運行下，能將顧客為主的精神完全融入服務流程，有效提昇顧客的滿意度，同時提昇員工的工作滿意度，造成服務品質提昇的正向循環。
研究結果顯示，研究所提出的機制透過與旅遊決策相關的顧客特性為預測基礎，有效的預測出顧客對不同服務人員的喜好，並指派合適服務人員進行服務。評估指標的表現上，人員推薦系統的平均準確率達81%，最高到85%。研究也證實，在未知服務人員服務特性的狀況下，所推薦出的服務人員，事後觀察其服務特性，發現與顧客所喜好的服務特性雷同，亦即利用此機制選擇服務人員，比起過往隨機指派服務人員的方式，更能滿足顧客對服務的期待。
另外，利用過去顧客的體驗服務經驗與當地的小費文化融合，計算出建議的服務小費金額給予顧客做參考。結果顯示，建議小費金額與顧客對該導覽人員的評分呈現正相關，亦即顧客的平均給分越高者，其小費建議金額會越高。證實了以此系統計算出的小費建議金額，能幫助顧客有所憑據的給予小費金額又不失自身感受到的服務價值。

關鍵字

服務小費 ； 推薦系統 ； 自我組織映射 ； 協同式推薦

並列摘要

The growth of service industry impacts the economic development nowadays. However, service heterogeneity is still one of the complex problems to maintain superior service quality. Existing researches attempted to discuss and solve the problem of unstable service quality caused by human beings, such as the appropriateness of job for service providers and education training of staff to standardize the process and control service quality. In addition, some literature investigated the role of customer in service delivering process and the gap between their expectation and perception. Nevertheless, a few researches emphasized on the effect of interaction between service providers and customers that may result in different level of service quality.
This research proposes a new service process by utilizing self-organizing maps and collaborative filtering to form a hybrid approach (including choosing the service providers and the amounts of given tips) based on customer perception. Through the unsupervised learning network, self-organizing maps can cluster and discover the similar segments of customers. Next, we use collaborative filtering approach to predict new customer’s preference based on similar segments.
The proposed approach can effectively forecast customer’s preferences among service providers and assign appropriate employee to serve. Based on customer service experience and the local culture for tips, we can calculate the appropriate amount of tips as recommendation for customers. To blend customer-oriented spirit into service process, the proposed method also can effectively improve the level of satisfaction of customers.
The result shows average value of MAP (mean average value) is 81% and the maximum value of MAP is 85%, which is good for the experiment. The attributes of recommending employees can fit to customer preference more. In other words, our approach can effectively bridge the gap between customer expectation and perception. Finally, the result also reflects on the amount of suggesting tips (i.e., the more the employee can match to the preference, the more tips customer may give).

並列關鍵字

Tips of service ； Self-organizing maps ； collaborative filtering approach ； Recommendation system

參考文獻

Hoffman, K. D. and T. N. Ingram (1992). Service provider job satisfaction and customer-oriented performance. Journal of Services Marketing, Vol. 6(2), pp. 68-78.

Ahn, H. J. (2008). “A new similarity measure for collaborative filtering to alleviate the new user cold-starting problem.” Information Sciences, Vol. 178(1), pp. 37-51.

Ambroise, C., Seze, G., Badran, F. and Thiria, S. (2000). Hierarchical clustering of self- organizing maps for cloud classification. Neurocomputing, Vol. 30, pp. 47–52.

Amold, S. J. (1979). A Test of Clusters. Journal of Marketing Reasearch, Vol. 16(4), pp. 545-551.

Barragans-Martinez, A. B., Costa-Montenegro, E., Burguillo, J. C., Rey-Lopez, M., Mikic-Fonte, F. A. and Peleteiro, A. (2010). A hybrid content-based and item-based collaborative filtering approach to recommend TV programs enhanced with singular value decomposition. Information Sciences, Vol. 180, pp. 4290–4311.