Identifying High Value Users in Twitter Based on Text Mining Approaches
Yibing Yang, M. Omair Shafiq · 2019
Finding out new potential users for specific products are always the needs of the marketing department in industries. While Traditional ways like RFM model perform poorly in exploring new users. While the popularity of social media like Twitter and Facebook provides advertisers a new way to find, understand and target their users. In this paper, we propose a new method to find out and rank high-value target audience for a specific brand by utilizing machine learning and text mining approach. Overall tweets from 10 accounts in Twitter are collected to build the target and non-target dataset. In order to solve the data imbalance problem, five data resampling methods are assessed. Ensemble learning approaches include Bagging and Boosting algorithm are used to build the classifier. The results show that SMOTE outperforms other resampling method and AdaBoosting algorithm outperform other single classifier and Bagging model. We also find out the existence of marking accounts exists so that a threshold is set to filter these accounts which are not real users. We believe that our approach could be used in industry for identifying high-value users for online marketing purpose.