以感性工學程序建構網頁設計系統之研究

A Study of Generating the Web Design System Based on the Kansei Engineering Process

目前網頁設計的相關研究眾多，但探討使用者對不同網頁設計的認知之相關研究卻不多見。所以，本研究以感性工學的程序，建構網頁設計系統。導入使用者的感性概念，探討何種網頁型態要素及感性語彙會影響使用者操作上認知。亦即為探討何種網頁的外觀會令人有不一樣的感受。本研究目的有三：（1）了解現有網頁設計狀況；（2）探討不同網頁對使用者操作的感覺；（3）研究結果可以提供日後網頁設計的參考。 本研究採用複迴歸分析與數量化一類等兩種線性模式與倒傳遞類神經網路非線性模式探討網頁形態與感性語彙之間的相關性，並安排六個新設計網頁，驗證線性分析模式與非線性分析模式差異為何？研究結果如下：（1）由數量化一類結果得知，七個型態項目中以〈版面編排〉、〈圖文比例〉與〈顏色量〉對『突出平凡』感性語彙之相關程度較大，而〈留白量〉與〈背景色彩〉之相關程度較小。（2）〈圖文比例〉要素對七組感性語彙之偏相關係數都高於0.7，此分數高於其他形態要素，意即〈圖文比例〉要素對網頁設計最重耍。其次為〈版面編排〉（除在『鄉土摩登』感性語彙外，其餘偏相關係數皆在0.7之上）；而以〈背景色彩〉對網頁感性影響較小。（3）三種模式中其中複迴歸分析與數量化一類模式之特質及調查結果兩者較為接近（預測相似度81%），而類神經網路與兩者之結果相去較遠（預測相似度分別為62%及52%）。

Although there are many studies about the design of Webs at present, the relative studies in exploring the user's cognition among the different Webs are not so popular. Hence, this study is based on the process of the Kansei Engineering to conduct the Web design system and introduce the Kansei concept to explore what style elements and Kansei vocabularies of Webs could influence the users' cognition when they access the Webs. In other words, this study is aim to investigates what's kind of the appearance of Webs could make users feel different. There are three purposes in this study. （1） To realize the current condition of the Web designs. （2） To investigate the feelings of different Webs that the users operate. （3） The results could be provided as the reference in Web design. This study adopts two linear models, Multiple Regression and Quantification Ⅰ, and Neural Network of non-linear model to investigate the relationship between the style elements and the Kansei vocabularies in Web designs. Further, Six new designed Webs were arranged for verifying the linear model or the non-linear model is different? The results show that : （1）According to the results of the Quantification Ⅰ model, the style elements of 'layout'、'proportion of picture-word' and 'number of colors used' are very related to the Kansei pair of 『Outstanding-Common』, but the 'quantity of blank space' and 'background color' are less relative. （2）The partial correlation coefficients are all above 0.7 between' proportion of picture-word' and seven Kansei pairs. The scores are larger than the other style elements. This means that 'proportion of picture-word' is the most important item for Web design. The second powerful item is the 'layout' （The partial correlation coefficients are above 0.7, except the 『Native-Modern』）; The least influential item is the' background color'. （3）The characteristics and performance are very similar between Multiple Regression model and Quantification Ⅰ model, but the Neural Network model quite differs the others.