Social Media Comment Management using SMOTE and Random Forest Algorithms
Nuanwan Soonthornphisaj, Taratep Sira-Aksorn, Pornchanok Suksankawanich · 2018
Comment posting is one of the main features found in social media. Comment responses enrich the social network function especially for how-to content. In this work, we focus on cooking video clips which are popular among users. Various questions found in comments need to be clustered in order to facilitate the clip owners to effectively provide responses for those viewers. We applied machine learning algorithms to learn and classified comments into predefined classes. Then the density-based clustering algorithm, DBSCAN, is applied to cluster the content of Comment. The experimental result show that using Random forest with SMOTE provides the best performance. We got 95% of the average performance measured in term of Fl-measure. Furthermore, we implement the incremental learning system via an online application that can automatically retrieve and organize video clip's comment into categories.