運用結構方程模型、熵、與模糊積分於模糊多評準決策－網路忠誠度模型之建構與測量

Abstract

網路忠誠度（電子商城忠誠度）能增進消費者於網路商店之再購意願及行為，並為電子商務服務公司創造利潤。然而，來自人類對於多準則之測量值通常是主觀認知的數據，大部份傳統測量方法不能在經過統計驗證之影響關係的基礎上，正確表示出不確定資訊的數值格式。因此，本研究目的是提出一種新的模糊多評準決策模型方法來整合電子商城忠誠度構念之概念化模型和測量模型。本研究提出之方法(SEM-FMCDME-FI法)能建構「消費者網路忠誠度」因果結構之概念模型，及演算忠誠度價值之測量值。SEM-FMCDME-FI法運用結構方程建構網路忠誠度的因果模型，並應用結構方程模型於模糊多評準決策問題，整合模糊語意尺度改善測量尺度，以CSCF演算法解模糊化，及採用資訊熵計算出企業間的多準則滿意度及忠誠度的競爭激烈程度指標，最後，以模糊積分法得出電子商城多準則交互作用之下的忠誠度測量值。本研究以三家台灣B2C電子商場零售服務網站調查實證模型可行性，實證研究顯示，增加電子零售商城忠誠度的關鍵前因準則為：交易安全、聯合促銷有用性、及商品推薦個人化。網路忠誠度之多準則資訊熵值結果顯示，公司間之網路忠誠度不確定性及差距最高的準則為：信譽口碑、搜尋速度、及購物愉快。模糊綜合效用測量值求解出消費者對於多公司多評準忠誠度優先排序之決策。經比較李克特量表、模糊加權法、模糊積分法之忠誠度值，結果顯示加法型及非加法型模糊積分測量值的價值排序不同。本研究的價值為提出基於結構方程模型的因果模型及驗證性因素分析，以改善模糊多評準決策之問題結構辨識階段之信度及效度﹔第二，應用模糊測度理論和以非加法型模糊積分可以演算出電子零售服務電子忠誠度之綜效效果測量值﹔第三，以模糊邏輯將消費者認知忠誠度轉換為測量數字﹔第四，資訊熵值作為競爭指標，可用以評估企業之間的準則忠誠度差距。網路忠誠度模型的建構及測量能協助電子商務零售服務企業建構出忠誠度績效，提供電子商務服務務創新的策略方向。本文提出之整合結構方程模型的模糊多準則決策技術，未來可供研究者應用於多方法論之研究，並支援經理人對於網路市場電子化服務之管理及創新。

Customer online loyalty (e-loyalty) facilitates repurchase intentions and behaviors at Internet stores and creates profitability for e-services firms. However, while the measured values of human perceptions toward various criteria are often subjective data, most conventional measurement methods can not precisely represent uncertain information by numerical format based on statistically-validated influential relations. The purpose of this study is to present a novel fuzzy multiple-criteria decision-making (FMCDM) modeling method (SEM-Fuzzy MCDM with Entropy and Fuzzy Integral (SEM-FMCDME-FI) to integrate the conceptualization and the measurement of the e-loyalty construct. We first used the structural equation modeling (SEM) for identifying the causal model of e-loyalty. We then extend this SEM causal model to a multi-criteria decision-making method (MCDM) model. Third, we applied the logic of triangular fuzzy numbers to measure consumer’s satisfaction levels and to transform these fuzzy data into explicit numbers into an evaluation matrix by linguistic variables scale, whose interval was quantified by consumers to replace the Likert scale. The Converting Fuzzy number to Crisp (CFCS) deffuzification algorithm converts fuzzy data into a crisp set. Next, we examine the competitiveness of e-loyalty criteria between firms by calculating the information entropy index. Finally, we adopted the fuzzy integral method to determine the interaction of multiple e-loyalty measures. An empirical study for evaluating three Taiwanese B2C e-commerce websites shows the feasibility of our proposed method. The results illustrate that main antecedents promoting e-retailing loyalty include security, usefulness of bundling promotion, and personalized product recommendations (PPRs). The loyalty entropy shows the criteria with the largest uncertainty between firms are: credibility, speed of search, and enjoyment. The fuzzy integral utility measures determine e-loyalty priority decision-making for the case companies. A comparison analysis of e-loyalty derived from additive FMCDM, non-additive FMCDM, and from Likert simple averaged weighted approach highlights the different ranking results. The proposed SEM-based relationship framework can statistically improve the validity and reliability of FMCDM modeling. The use of information entropy provides a reduced complexity for assessing the e-loyalty criteria gaps between firms. Moreover, application of fuzzy measure theory and fuzzy integral can represent the e-loyalty synergies gained from decision-makers for e-retailing services. The proposed SEM-enabled FMCDM techniques can be an effective approach for researchers in multi-methodology studies, and for practitioners in supporting management of e-services innovation in online markets.