E-Commerce Web Site Trust Assessment Based on Text Analysis

Banatus Soiraya ； Anirach Mingkhwan ； Choochart Haruechaiyasak

《International Journal of Business and Information》 3卷1期 (2008/06) Pp. 86-114

https://doi.org/10.6702/ijbi.2008.3.1.4

並列摘要

This paper reports the efficiency and effectiveness of our proposed text analysis module for a Web-site trust assessment system. One main factor that influences the level of trust for an e-commerce Web site is the textual content appearing on each Web page, especially the main page. The textual content refers to words, phrases, and sentences appearing on the page. A Web site that has a high trust level should contain meaningful keywords related to the e-commerce domain, such as return policy, payment option, and security. To analyze the textual content, our text analysis module adopts automatic classification techniques to learn from the example Web site data set. We perform experiments on two ecommerce domains (a jewelry store and a book shop) by using three well-known classification algorithms. A sample set of Web sites under each domain are collected and labeled as trust or untrust for performing our experiments. Two approaches for constructing the feature set are (1) using all extracted words from the textual contents (baseline), and (2) mapping extracted words into the meaningful groups of e-commerce terminology (EC-word). The best text analysis result of 83.5% accuracy was obtained when the support vector machines based on a sequential minimal optimization (SMO) algorithm was applied with the EC-word feature set.

並列關鍵字

e-commerce ； trust assessment ； Web mining ； text classification