Optimizing Feature Selection Enhancing Sentiment Analysis With Fxtend Algorithm

Abstract

Student sentiment analysis is the process of analyzing the feelings, opinions, and attitudes of students towards various aspects of their educational experience. This study proposes a feature selection method utilizing the FXtend algorithm to enhance sentiment analysis tasks. The approach involves employing Recursive Feature Elimination (RFE) with three distinct classifiers: ElasticNet, Extra Trees Classifier, and Gradient Boosting Classifier. Through iterative elimination, less relevant features are identified, aiming to retain the most informative ones for sentiment analysis. Subsequently, sentiment scores are assigned to each token in pre-processed text based on the selected features, Parts of Speech (PoS) tags, and the presence of opinion words. Aggregating these scores provides an overall sentiment assessment for the text or document. Finally, sentiment scores are normalized to a standardized scale, facilitating better interpretability and comparison across texts. This methodology promises improved accuracy and efficiency in sentiment analysis tasks, aiding in extracting meaningful insights from textual data.