Towards Automatic Online Persona Classification: A Comparative Study of Machine Learning and Deep Learning Models

The advent of internet causes inventions of social media, in which are applications that relied on user generated contents (UGC). These massive increases in social media users may help businesses and institutions in understanding and classifying their potential users or understanding which institutions a user’s belongs to help in profiling. This research is using X (formerly twitter) followers’ posts from 4 of famous universities in Indonesia. This research compares several machine learning and deep learning models and incorporates several approaches in handling overfitting. The best model we 1D CNN for both non balanced data and undersampled data (72% f1-score and 73% f1-score respectively). However, using undersample and dropout didn’t improve the overfitting problems.