Developing an Intelligent System with Deep Learning Algorithms for Sentiment Analysis of E-Commerce Product Reviews.

Area of Science:

• Natural Language Processing
• Machine Learning
• E-commerce Analytics

Background:

• Online reviews significantly influence consumer purchasing decisions in e-commerce.
• Deceptive reviews pose a persistent challenge, undermining the reliability of online feedback.
• Accurate analysis of consumer sentiment is crucial for businesses to refine market strategies and understand brand perception.

Purpose of the Study:

• To develop and evaluate deep learning models for sentiment analysis of e-commerce reviews.
• To accurately classify consumer sentiment as positive or negative, addressing the issue of deceptive reviews.
• To leverage opinion mining for enhanced business intelligence and product development.

Main Methods:

• Utilized Long Short-Term Memory (LSTM) and Convolutional Neural Network integrated with LSTM (CNN-LSTM) deep learning models.
• Applied data preprocessing techniques including lowercasing, stopword removal, punctuation removal, and tokenization for data cleaning.
• Trained and tested models on real-time e-commerce product reviews from Amazon, covering various electronic categories.

Main Results:

• The LSTM model achieved an accuracy of 94% in sentiment classification.
• The CNN-LSTM model demonstrated an accuracy of 91% in classifying consumer sentiment.
• Both deep learning models proved effective in detecting and classifying sentiment in e-commerce reviews.

Conclusions:

• Deep learning techniques, specifically LSTM and CNN-LSTM, offer optimal performance for classifying customer sentiment in e-commerce.
• The proposed system provides valuable insights into consumer opinions, aiding businesses in strategic improvements.
• Sentiment analysis of online reviews using advanced AI is a viable solution for understanding customer feedback and combating deceptive practices.