In order obtain all kinds of home textile websites' key elements, in this paper, we design the corresponding questionnaires for the portal-type. comprehensive-type, profession-type and brand-type home textile website, and use the method support vector machine to classify and predict the questionnaire data by adjusting the number of training parameters. the number of the training parameters is selected from questionnaire options which could be reduced one by one untill we get the optimal prediction result. the rest of training parameters would be the webpage's key elements which are satisfied the customer most. trough training, we found that color, services and help, recommed information and related rankings and news information play important roles when customers buy some product in the professional websites; logo, font, color, pictures and background, product information, recommed and discounts information, service and help, famous brand, advertising position, commodity richness, promotion, the user evaluation and purchgase, all these numerous factors of brand-type website are influential on customer statisfaction; font, color, organization structure and information density, navigation, recommed and discount information and advertising proportion are the main factors of comprehensive home textile website; navigation, information classification, search, links, promotion and propaganda, information, famous brand, service and help, exhibitions and business information to users in portal-type website evaluation have crucial effect. all these result can be used to the future home textile website construction.