Hybrid Feature Selection on Social Media Dataset for Sentiment Classification using Deep Learning Techniques

Abstract

Sentiment classification involves determining the sentiment expressed in text, such as positive, negative, or neutral, but social media data presents challenges due to its high dimensionality, noise, and unstructured nature. This study proposes a novel sentiment classification approach by combining hybrid feature selection methods with deep learning techniques. Social media platforms generate vast amounts of data daily, which is often noisy, redundant, and irrelevant for sentiment analysis. Hybrid feature selection techniques, which integrate filter and wrapper-based methods, assist in reducing the feature space while retaining the most informative features. By applying deep learning models, such as recurrent neural networks (RNNs) and long short-term memory (LSTM) networks, classification performance can be substantially enhanced. The proposed framework uses hybrid feature selection to eliminate noisy and irrelevant features, thereby improving the model's generalization capabilities. Experimental results reveal that the combination of hybrid feature selection and deep learning techniques not only boosts sentiment classification accuracy but also decreases computational overhead. This study highlights the effectiveness of merging traditional feature selection methods with modern deep learning models to better address the complexities of social media datasets and deliver more precise sentiment analysis. The results achieved by proposed model is 98.50% on social media dataset which is higher than conventional approaches.