Publication

Abstract : n this paper, we present a study on socialmedia data analysis of Malayalam YouTubecomments, specifically focusing on sentimentanalysis and emotion detection. Our researchaims to investigate the effectiveness of vari-ous machine learning (ML) and deep learning(DL) models in addressing these two tasks. Forsentiment analysis, we collected a dataset con-sisting of 3064 comments, while for two-classemotion detection, we used a dataset of 817comments.In the sentiment analysis phase, weexplored multiple ML and DL models, includ-ing traditional algorithms such as Support Vec-tor Machines (SVM), Naive Bayes, K-NearestNeighbors (KNN), MLP Classifier, DecisionTree, and Random Forests. Additionally, weutilized DL models such as Recurrent NeuralNetworks (RNN),Long Short-Term Memory(LSTM) and Gated Recurrent Unit(GRU). Toenhance the performance of these models, wepreprocessed the Malayalam YouTube com-ments by tokenizing and removing stop words.Experimental results revealed that DL mod-els achieved higher accuracy compared to MLmodels, indicating their ability to capture thecomplex patterns and nuances in the Malay-alam language.Furthermore, we extended ouranalysis to emotion detection, which involveddealing with limited annotated data. This taskis closely related to social media data analy-sis. For emotion detection, we employed thesame ML models used in the sentiment analy-sis phase. Our dataset of 817 comments wasannotated with two emotions: Happy and Sad.We trained the models to classify the commentsinto these emotion classes and analyzed the ac-curacy of the different models