Evaluate the Sentiment Analysis Performance of Several Classifiers Using Ensemble Feature Selection Method

Abstract: The growing availability of digital text data has sparked a need for effective sentiment analysis methods, which enable the automatic extraction of sentiment from text for various purposes. This study investigates the application of sentiment analysis using the Amazon Reviews Polarity Dataset, a curated compilation of texts categorised into positive and negative attitudes based on review ratings. Despite the dataset containing a substantial amount of labelled data, its narrow focus and classification technique have difficulties in effectively capturing nuanced expressions of sentiment. Using the Amazon Reviews Polarity Dataset as a foundation, this work does a thorough investigation of the performance of several machine learning models for sentiment analysis tasks. We learn a lot about how well different models can detect sentiment patterns by comparing them using a wide variety of performance metrics like recall, accuracy, precision, and F1 score. These metrics include K-Nearest Neighbours (KNN), Random Forest (RF), Logistic Regression (LG), and Ensemble Classifier (ECLF). Logistic Regression stands out as the most effective model, with the Ensemble Classifier coming in a close second, highlighting their promise for practical sentiment analysis applications. These findings emphasise the usefulness of the dataset and the success of the models, showing the crucial significance of careful model selection and evaluation methods in guaranteeing dependable and accurate outcomes in the field of natural language processing applications