近年來資訊科技日新月異,人類的生活方式也逐漸改變,變得更簡單、更直覺性。「電子書」相關議題也成為讀者間討論的重點,究竟是「電子閱讀」或「紙本閱讀」何者較為優勢,但其結果反應出,未來閱讀書籍將不再侷限於印刷形式的傳統書本,而將書本內容數位化或電子化,透過不同的載體供使用者閱讀與利用。因此,本研究將預測並評估消費者對於電子書之方案屬性變數、社經變數、生活型態變數與購買因素等變數下,能得到更精確研究結果以及瞭解消費者實際選擇購買電子書行為之效用。 本研究以類神經網路探討消費者於電子書與實體書狀況下實際選擇行為預測模式之建構,是否優於二項羅吉特實際選擇模型,分別以兩模型之錯誤分類表,判斷模型間正確預測能力。然而,將模型預測結果,透過「消費者每月購書預算」之問項統計予以市場區隔;並將兩種模型,分別對高購書預算與低購書預算兩群體,進一步驗證預測模型的準確性。 最後,本研究將使用市場區隔結果分別以倒傳遞神經網路進行變數間敏感度分析篩選出顯著變數,以瞭解其不同的目標「族群」結構、對於消費市場所帶來的影響性及喜好程度之敏感程度,進而判斷電子書項目影響消費者之因素,以提供學者後續研究及出版業者未來研發電子書商品之參考依據。
As information technology is rapidly changing in these recent years, the living ways of humans have gradually changed and becoming more simple and intuitive. Therefore topics related to the “E-book” have become the focus in the discussions among readers. Regardless whether it is “E-reading” of “paper reading” having the advantage, the results show that future reading shall not be only confined in the traditional printing form of books. On the contrary, the book contents are digitalized or electronized, and then read or used by the user through different platforms. Because of this future trend, this study shall predict and assess the factors, such as variable attributes, social-economic variables, life style variables and purchase factors, of E-book projects. The research is to obtain a more accurate result to understand the effects toward consumers’ behavior when actually purchasing E-books. In this study, the artificial neural network was applied to discuss the actual selecting behaviors of the consumer toward E-books and real (paper) books, and to further construct a prediction model in purchasing behavior. The model was tested determine whether it was better than the binary logit model: both models were evaluated via error classification, respectively, to determine their accuracy in prediction. The results of the two models were further specialized based on the survey in “monthly budget of purchasing books” for market segmentation; two groups, high budget and low budget in purchasing books, were applied to further verify the accuracy of the prediction models. Finally, the research used the market segmentation results to achieve the significant variables by applying the back-propagation network respectively to the two models for sensitivity analysis. The obtained significant variables enables us to understand how the different objectives from “group” structures bring influence, favorability and sensitivity to the market, and to further determine how the E-book factor affects consumers. The results in this study may serve as reference for future research and publishers in developing E-book merchandises.