The Study on Prediction Models of Social Network Involvement Degree in Taiwan

• Main Authors: Ching-Ting Hsu, 許景婷
• Other Authors: Te-Hsin Liang
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/28671829246390498815

碩士 === 輔仁大學 === 應用統計學研究所 === 99 === This study examines the differences of involvement degrees of social network service in terms of demographic profile, internet case, broadband case, wireless mobile internet and broadband service of internet users. Data were collected in Taiwan Network Information Center (TWNIC) using Taiwan Broadband surveys from 2010. The sample size of Social network users’ data as target research variable was 1,049. Important variables were selected by testing the significances between independent variables and target variable, then dealing with the multicollinearity among independent variables. By using Multiple Logistic Regression, this study built a whole model with all training dataset and three kinds of ensemble models which are to calculate the average, the maximum and the vote of the predictive probabilities or results from 100 Bootstrap models. The result shows that the best model is “The Whole Model” and significant variables are "sex", "work", "broadband network fees_home", "broadband network_chat" and "broadband network_space".