Optimized Deep Learning Model for Sentimental Analysis to Improve Consumer Experience in E-Commerce Websites

Abstract

Sentiment analysis plays a pivotal role in deciphering customer sentiments from vast amounts of unstructured data, particularly in the context of e-commerce where customer reviews are prolific. The evolution of e-commerce reviews toward a multimodal format, including images, videos, and emojis, introduces new dimensions to sentiment analysis. Traditional text-based models may struggle to effectively capture sentiments expressed through non-textual elements. This paper proposed an effective sentiment analysis model for the E-Commerce Platform to improve the user consumer experience. The proposed method comprises Fejer Kernel filtering for data points estimation in the E-commerce dataset points. Within the estimated data points fuzzy dictionary-based semantic word feature extraction is performed for the estimation of features in the E-Commerce dataset. The dataset for the analysis is computed with the Optimized Stimulated Annealing for the feature extraction and selection. The classification of customer opinion is classified with the BERT deep learning model. The feature extracted from the model is the opinion of consumers in the E-Commerce dataset. The classification of consumer preference experience is based on opinion of customers in the E-commerce dataset. Simulation results demonstrated that proposed model achieves the higher classification accuracy for the E-Commerce platform.