Identification of Salient Attributes in Social Network: A Data Mining Approach

Abstract

Social Media is a source with tremendous volumes of data which is only growing by the day. Taking a cue from earlier studies on data generated from social media, this study is based on a dataset originally drawn from the Facebook social network page of a large multinational cosmetics company. A total of 500 out of 790 posts published on the social media page were analyzed through the data mining classification technique i.e. linear regression. Nine numeric attributes (variables) were regressed with one attribute considered as the criterion attribute and the rest eight as predictor attributes. The eight dependent variables were: V1 “lifetime post total reach”, V2 “lifetime post total impressions”, V3 “lifetime engaged users”, V4 “lifetime post consumers”, V5 “lifetime post consumptions”, V6 “lifetime post impressions by people who have liked your page”, V7 “lifetime post reach by people who like your page”, V8 “lifetime people who have liked your page and engaged with your post”; and the independent variable was V9 “Total interactions - the sum total of comments, likes, and shares of a post”. The results show that not all predictors are significant in explaining the criterion variable. Similarly, a correlation matrix was generated where the inter-attribute correlation among all nine attributes was calculated. The results of association drawn from correlation are different from regression which depicts the fundamental difference and approach of these two techniques. WEKA version 3.8 was the data mining software used to analyze the dataset.