Datasets details.

Sentiment analysis is a pivotal domain in Natural Language Processing (NLP), particularly for understanding opinions expressed in sequential and textual data with the usage of machine learning. It involves identifying and categorizing emotions expressed in textual reviews and messages. Social media platforms such as Twitter, Facebook, and Instagram generate extensive datasets rich in sentiments, making their analysis crucial for monitoring public opinion and informing business strategy. By uncovering customer satisfaction levels, product feedback, and service-related concerns, sentiment analysis helps organizations refine marketing efforts, optimize product features, and improve service delivery. Traditional machine learning techniques struggle to process large datasets and yield accurate results efficiently. To address this, we propose an effective multi-layer perceptron deep network with word embedding features, called MultiSentiNet, for sentiment analysis on Twitter datasets. The proposed model’s performance is evaluated against conventional machine learning classifiers and state-of-the-art deep learning classifiers, indicating superior accuracy with three different datasets. The significance of the proposed model is further tested on three diverse datasets (women’s e-commerce, US airline sentiments, and hate text-speech detection) that demonstrate that the proposed framework outperforms other classifiers in terms of accuracy, recall, precision, and F1 score. The performance of the proposed model is compared with previously published research works. Furthermore, the interpretability and analysis of MultiSentiNet results are explained using the LIME XAI technique, providing deeper insights into the model’s predictions and practical value in strategic business decision-making.