Class imbalanced detection of sarcasm in social networking websites based on minority up-weighting
Rahul Sen, Mehtab Alam Khan, Md. Tarik Aziz · Journal of Emerging Technologies and Innovative Research · 2021
Over the past few decades, the evolution of machine learning has been tremendous, which gave solutions to many real-world problems and still making lives more comfortable as well as innovative. One of the applications of machine learning would be to detect sarcasm in social media like Twitter, Facebook etc., which will, in turn, give us insight into a topic trending in social networking websites. Thus this mining of people’s opinions would be of use to make decisions in many ways. But due to the highly imbalanced classes of social media data, it is often difficult to get the accuracy that is desired. So, to deal with this type of dataset, in this study, synthetic minority oversampling based methods are proposed. The main focus of this article is minority up-weighting, hence, two methods namely KMeansSMOTE and BorderlineSMOTE algorithm are used along with classifiers like Bernoulli Naive Bayes (BNB), Multilayer Perceptron (MLP) and Decision Tree Classifier (DTC) to get better accuracy while dealing with imbalanced dataset. Thus this article contains an analysis of imbalanced classification in the detection of sarcasm in social media through minority up-weighting techniques.