Volume 12, Issue 9

Generalized Study on Sentiment Analysis on Social Media Using Machine Learning Techniques

Author

Mohamed Bodea*1 and Dr. İsmail Yıldız2

Abstract

This paper discusses the increasing interest in using machine learning techniques to analyze human-written texts due to their widespread availability. Natural Language Processing (NLP) is essential for extracting accurate knowledge from these texts, but it faces challenges due to the vast amount of information and complex relationships among words. The proposed study introduces a novel feature selection technique using sentiment-based word vectors and knowledge about word influence on classifiers. An artificial neural network trained with reinforcement learning is employed to assess the impact of removing each word from the training dataset, using word embeddings for representation. This method predicts word ranks efficiently without complex statistical computations, even for new words in the corpus. The evaluation of the proposed method demonstrates its high accuracy (94.61%) in predicting word ranks not included in the training set. Additionally, feature selection based on these ranks enhances the performance of various classifiers, including Support Vector Machine (SVM), Naïve Bayes (NB), Random Forest (RF), Convolutional Neural Network (CNN), Long-Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU). Among these classifiers, the GRU classifier achieves the highest accuracy (95.54%), surpassing other classifiers and state-of-the-art methods in literature.

DOI

https://doi.org/10.62226/ijarst20231127

PAGES : 0 | 0 VIEWS | 0 DOWNLOADS

Download Full Article