Unveiling sentiment analysis: A comparative study of LSTM and Logistic regression models with XAI insights

Abstract

In this study, we delve into sentiment analysis and the role of Explainable Artificial Intelligence (XAI), with a focus on techniques such as Lime that bring transparency to machine learning (Logistic Regression) and deep learning (LSTM) models. We explore how ML predictions can be biased using XAI and how XAI helps us understand DL models used in sentiment analysis through research that has been made. Examining various research, we notice a gap – the lack of training and interpretation for both ML and DL models on the same dataset using XAI. Our research fills this gap, shedding light on ML and DL model predictions through XAI's lens. By completing our research work, we come to know that even with an accuracy level of 83% for the DL model, they outperform the ML model with an accuracy level of 92% in some cases. This distinction is only identified with XAI techniques, particularly Lime.