以類神經網路探討網頁視覺圖像複雜度偏好之研究

The Study of User's Favor on the Complexity of Vision for Web Pages Using Neural Networks

本研究以三個虛擬網頁視覺圖像為測試樣本，分別以高複雜度、中複雜度、低複雜度三種不同複雜程度的網頁視覺圖像，測試十一個不同年齡層的受試者，以分層隨機抽樣法，不分性別，排除網頁文字因素、網頁色彩因素與其他會干擾網頁視覺感受之相關因素，做網頁視覺圖像複雜度的偏好測試，將問卷調查的統計結果，與應用類神經網路模型運算所得的結果比較。類神經網路模型的進行步驟為：先輸入單數組之年齡組，經由類神經網路的學習與預測得知結果。利用單數組之年齡組的偏好百分比數據，去推論雙數組的年齡組對網頁視覺圖像複雜度偏好的百分比數據，將類神經網路模型所得到的結果，與問卷調查的統計結果相互驗證，結果證明以類神經網路模型運算，實可以用較少的樣本數來測試，得到與較多樣本數測試的結果相似，且其誤差小於5%，驗證本研究運用類神經網路之RBF-NN模型實驗的「合理性」及「價值性」。本研究之結果顯示：中複雜度的網頁視覺圖像，普遍來說接受度較高，年齡層較高的受試者對高複雜度的網頁視覺圖像較偏愛。

In this paper, the age effects on the favor of the visual complexity of web pages are studied using the model of neural networks. The visual complexity is divided into high complexity, middle complexity, and low complexity. The age of the tested people is from 11 to 65 and is equally divided into 11 groups with each group 40 tested people. The effects of sex, words, colors, and other visual factors are ignored. With the use of neural networks, 6 groups of tested people are used as the training data sets. This trained neural network is then used to predict of tendency of the total 11 groups of tested people toward the favor of the visual complexity of web pages. Experimental results show that the trained neural network can predict accurately. Furthermore, the web pages of middle complexity are popular with most people, and those of high complexity are popular with the old people.