Enhancing E-commerce recommendations with sentiment analysis using MLA-EDTCNet and collaborative filtering.

Area of Science:

• Artificial Intelligence
• Data Science
• E-commerce Technology

Background:

• E-commerce growth necessitates improved product recommendation systems.
• Current systems often struggle with accuracy and user satisfaction.
• Integrating user sentiment with traditional recommendation methods is key.

Purpose of the Study:

• To develop an advanced recommendation framework integrating sentiment analysis (SA) and collaborative filtering (CF).
• To enhance recommendation accuracy, precision, recall, F1-score, and AUC.
• To improve overall user satisfaction and engagement in e-commerce platforms.

Main Methods:

• Feature-level sentiment analysis using a Multi-Layer Attention-based Encoder-Decoder Temporal Convolution Neural Network (MLA-EDTCNet).
• Log-term frequency-based modified inverse class frequency (LFMI) for feature extraction.
• Modified Conditional Generative Adversarial Network (MCGAN) for class imbalance and Ocotillo Optimization Algorithm (OcOA) for parameter tuning.

Main Results:

• The proposed framework significantly outperforms state-of-the-art models across key performance metrics (accuracy, precision, recall, F1-score, AUC).
• The integration of SA and CF provides highly personalized product recommendations.
• Demonstrated effectiveness in addressing class imbalance and feature extraction challenges.

Conclusions:

• Hybrid deep learning techniques, combining SA and CF, offer a robust solution for e-commerce recommendation systems.
• The framework enhances user experience through accurate and personalized recommendations.
• This approach holds significant potential for driving business success in online retail.