行動電子商務之多維度使用偏好分析

Abstract

本論文探討不同國家的行動電子商務使用者之多維度偏好分析。為判別從不同國家取得的樣本是否存在差異，本研究提出一個基於資訊複雜度理論的新方法來進行參數模型的選擇及不同維度的變數分類。依據資訊複雜度理論所提供的參考值，可決定來自不同樣本的資料是否能以性質相同的參數模型加以描述。相似的方法將用以判別在先驗的參數模型中，不同維度內的個別變數解釋模型差異程度之能力。藉由本研究所提出的新方法判定參數模型並進行不同維度之變數分類，將可鑑別不同國家的行動電子商務使用者之偏好，以為建立相應統計模型的基礎。 灰關聯分析亦可用於判別使用者偏好，因此本論文也將探討應用這兩種方法於判別使用者偏好之特性。基本上，這兩種方法都能顯示不同樣本間是否存在差異。依據灰關聯理論的方法見長於樣本內之偏好判定，而依據資訊複雜度的方法則可提供建構最適參數模型之依據。本論文能為學術界提供兩方面的貢獻，其一為基於資訊複雜度理論的新分析方法，另外則是這個新方法與灰關聯分析的特性比較。 使用行動電子商務之服務，基本上可視為一種選擇行為的問題，由於本論文專注於行動電子商務的使用行為分析，特別是其相應的分析架構，因此由作者所倡議的相關方法，亦將一併介紹於本論文中。這些方法包括應用模糊德爾非法與狩野模型於形成電子市集關鍵因素的分析。這些進程可拓展未來相關研究的視野，且所提供的建議將指引更具體的研究方向。

This study is committed to conduct the multi-dimensional preference analyses for the diverse usage in m-commerce among the countries. In order to determine whether there is a discrepancy between samples, this study proposed a new approach based on ICOMP to perform model selection and variables clustering in different dimensions. According to the ICOMP criteria, the model selection may imply that there is heterogeneity between samples. The variable clustering, whatever within different dimensions, may indicate that there are differences in ranking the most relevant variables between samples. That is, results based on the proposed approach could be used to support that there were discrepancies in using m-commerce services among the countries, whatever in model selection or variables clustering. Characters of both the ICOMP approach and the grey relational analysis in distinguishing the diverse usage would be discussed. Basically, both approaches can conclude there are discrepancies between the samples. The grey relational analysis is good to determine the preferences within a sample; and the ICOMP approach can determine the best fitting models for samples. The contribution of this work towards the field in two folds: the new approach to analyze the discrepancy; and the comparison between the two approaches. Basically, the usage of services in m-commerce can be treated as a choice behavior problem. Since this dissertation is dedicated to the analysis of usage in m-commerce, especially the framework for the analyses, connected methodologies which were proposed by the author during the development were also introduced, e.g. applying Fuzzy Delphi as well as the Kano model for the key factors analyzing in forming an e-marketplace. The progress may benefit further research by broadening its scope and suggestions made for more promising research directions.