Word Embedding Feature for Improvement Machine Learning Performance in Sentiment Analysis of Disney Plus Hotstar Comments

Aug 4, 2026

Upload by Asher

This research applies machine learning methods combined with word embedding features to process social media comments from the Disney Plus Hotstar application. Word2Vec, GloVe, and FastText are used as embedding representations to test their effects on classification performance. Naive Bayes, K-Nearest Neighbor, and Random Forest are evaluated, considering each method’s known limitations and sensitivity to feature selection. Testing results show KNN reaches the highest accuracy before and after feature selection, while FastText yields the strongest overall KNN performance across balanced accuracy, precision, recall, and F1-score.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

Introduction

Sentiment analysis and social media data

Word embeddings and feature engineering

Machine learning classifiers and limitations

Research contribution and experimental focus

Word Embedding Feature for Improvement Machine Learning Performance in Sentiment Analysis Disney Plus Hotstar Comments Jasmir, Nurhadi, Eni Rohaini, M Riza Pahlevi, Daniel Sintong Pardamean Simanjuntak Departement of Computer Engineering, Faculty of Computer Science, Universitas Dinamika Bangsa, Jambi, Indonesia ARTICLE INFO Article history: Received May 09, 2024 Revised June 25, 2024 Published July 02, 2024 Keywords: Text Classification; Machine Learning Evaluation; Word Embedding; Sentiment Analysis; Social Media Analysis Corresponding Author: ABSTRACT In this research we apply several machine learning methods and word embedding features to process social media data, specifically comments on the Disney Plus Hotstar application. The word embedding features used include Word2Vec, GloVe, and FastText. Our aim is to evaluate the impact of these features on the classification performance of machine learning methods such as Naive Bayes (NB), K-Nearest Neighbor (KNN), and Random Forest (RF) . NB is very simple and efficient and very sensitive to feature selection. Meanwhile, KNN is known for its weaknesses such as biased k values, overly complex computations, memory limitations, and ignoring irrelevant attributes. Then RF has a weakness, namely that the evaluation value can change significantly with just a slight change in the data. Feature selection in text classification is crucial for enhancing scalability, efficiency, and accuracy. Our testing results indicate that KNN achieved the highest accuracy both before and after feature selection. The FastText feature led to the highest performance for KNN, yielding balanced accuracy, precision, recall, and F1-score values. This work is licensed under a Creative Commons Attribution-Share Alike 4.0 Jasmir, Departement of Computer Engineering, Faculty of Computer Science, Universitas Dinamika Bangsa, Jln. Jendral Sudirman, Tehok, Jambi Selatan, Jambi, Indonesia, Email: [ijay_jasmir@yahoo.com](ijay_jasmir@yahoo.com) 1. INTRODUCTION The integration of data derived from social media represents a significant advancement, offering an alternative data source to traditional data collection methods [1], [2], [3] . Social media data collection is efficient in various aspects, including cost-effectiveness, real-time data acquisition, and the ability to capture detailed community opinions [4], [5] . The analysis of public responses and opinions using social media data is known as sentiment analysis [6], [7], [8] . Sentiment analysis, a subset of natural language processing (NLP), employs machine learning methods to identify and extract factual details and emotional nuances from written text, determining the general sentiment—positive, neutral, or negative—expressed by the writer [9], [10], [11] . Applying sentiment analysis to extensive textual datasets, such as social media updates or user comments, allows for comprehensive analysis of public sentiment [12], [13], [14] . Several previous studies have explored sentiment analysis. For instance, Elik Hari Muktafin analyzed public service customer satisfaction using KNN with the TF-IDF feature, achieving an accuracy of 74%[15] . Heru Agus Santoso et al. examined sentiment analysis of hoax news using Naive Bayes, resulting in an accuracy of 77%[9] . Meanwhile, M. Ali Fauzi conducted sentiment analysis in Indonesian using Random Forest, achieving an average Out of Bag (OOB) value of 82,9%[16] . Ari Basuki conducted research on Sentiment Analysis of Customer Reviews of Delivery Service Providers on Twitter Using Naive Bayes Classification and produced a low accuracy of 50.6%[17] . Kartikasari Kusuma Agustiningsih analyzed Indonesian public sentiment towards the COVID-19 vaccine on Twitter using BiLSTM and word embedding features, namely FastText and Glove. The combination of BLSTM and FastText produces an accuracy of 75.76%. The combination of BLSTM and GLove produces an accuracy of 74.70%[17] . These studies indicate

FAQs

Which word embedding features are tested for sentiment classification?

The study uses Word2Vec, GloVe, and FastText as word embedding features to represent text for classification.

How do the evaluated classifiers perform, and which one reaches the highest accuracy?

K-Nearest Neighbor (KNN) achieves the highest accuracy both before and after feature selection in the experiments.

What embedding feature leads to the best KNN performance metrics?

FastText produces the highest performance for KNN, delivering strong and balanced values for accuracy, precision, recall, and F1-score.

Aug 4, 2026

Upload by Asher