Two important factors which indirectly influence the Internet shoppers to make some online purchases are the visual layout and the presentation of web page. In this paper, we propose an approach of web page layout analysis in order to assess the design of e-commerce web sites. Firstly, our proposed method segments each web page into five different blocks: top, left, center, right and bottom. We study three main feature types: spatial, location and presentation, which effect the design of each page. The spatial feature set includes the width-length ratio and relative area of each block while the location feature set is the appearance positions of e-commerce components such as navigation, product index and customer service.

The presentation feature set represents page characteristic and navigation menu format that appear on the page or block such as navigator style, menu type, image alignment and text density. To perform the experiments, we consider the top 100 Internet retailers as ranked by the Internet Retailer’s Top 500 Guide1 . We adopt classification algorithms in order to construct a model from the top web sites against other 100 randomly selected e-commerce web sites. The experimental results showed that the model which is constructed based on the presentation feature set yields better performance than using the spatial and location feature sets.

Previous study showed that visual and layout are two important influencing factors on web page design [3], [4], [7].Customers may stay or leave those web sites depend on the first impression with the appearance on the web page. Existing work has been performed to study some features of web page format and location appearance on the screen. Adkission, H. et al. [1] studied the location occurrence of e-commerce web sites component but did not perform to the other features, nor identify the good design in the e-commerce web sites.

Ivory, M.Y. et al. [3] only identified the general web site pattern but did not consider the good design pattern in e-commerce web sites. Robins, D. et al. [7] illustrated that high aesthetic treatment in the web sites will present a professional look and feel appropriate for organization. Their result has been done by human judgment. Kovacevic, M. et al. [5] considered web page segmentation to five blocks for improving search engine techniques.

Their results showed better performance when using web page segmentation compare with un-segmentation. Cai, D. et al. [2] implemented VIPS algorithm for segmentation web page by using Dom tree technique and visual format in the web page. Dom tree is initially introduced for representation HTML object rather than semantic or visual meaning. Wu, C. et al. [9] adopted VIPS algorithm for extraction of the product information of e-commerce web sites.

Open chat