Sustainable sentiment analysis on E-commerce platforms using a weighted parallel hybrid deep learning approach for...

Area of Science:

• Natural Language Processing (NLP)
• Machine Learning (ML)
• Deep Learning (DL)

Background:

• E-commerce platforms generate vast amounts of user reviews.
• Understanding customer sentiment from these reviews is crucial for improving customer satisfaction.
• Traditional methods struggle with the nuances of human language in reviews.

Purpose of the Study:

• To introduce a novel Weighted Parallel Hybrid Deep Learning-based Sentiment Analysis on E-Commerce Product Reviews (WPHDL-SAEPR) system.
• To accurately distinguish between positive, neutral, and negative sentiments in e-commerce product reviews.
• To enhance the effectiveness of sentiment analysis for e-commerce applications.

Main Methods:

• Utilized deep learning (DL) for its ability to capture complex linguistic relationships and implied emotions.
• Employed word2vec for word embedding and a novel WPHDL model for sentiment classification.
• Integrated Restricted Boltzmann Machine (RBM) and Singular Value Decomposition (SVD) within the WPHDL model.
• Implemented data pre-processing techniques to ensure compatibility and accuracy.

Main Results:

• The WPHDL-SAEPR approach demonstrated accurate sentiment classification on a consumer review database.
• The system effectively identified and distinguished various sentiments expressed in online reviews.
• The simulation results highlighted the performance improvements at each stage of the WPHDL-SAEPR architecture.

Conclusions:

• The proposed WPHDL-SAEPR system offers a robust and accurate solution for sentiment analysis in e-commerce.
• This method can significantly contribute to understanding consumer attitudes and enhancing customer happiness.
• The integration of DL techniques, RBM, and SVD provides a powerful framework for analyzing subjective data.