Knowledge and Information Systems Human-Computer Interaction

Fuzzy quantum-driven sentiment-aware recommendation system using Hamiltonian quantum generative self-attention Dandelion network

Narayanappa, Geetha, Jagannath, Girija
Springer Nature · 2026-11-01 · DOI 10.1007/s10115-026-02896-x

Abstract

With the rapid expansion of the online business, vast amounts of diverse customer data have been generated, making it difficult to leverage customer preferences efficiently, perform sentiment analysis from product reviews, and accurately recommend products. The accuracy of current recommendation systems is decreased due to issues with cold start, data sparsity, and the complexity of the linguistic expressions in customer evaluations. The research introduced the method named Hamiltonian quantum generative self-attention dandelion network (HQGSD-Net), an integrated approach that incorporates five different datasets for comprehensive sentiment and behavioral analysis. The data are pre-processed by using quality-driven fuzzy min–max neural networks (QDFMM Net) with tokenization, POS tagging, and parsing to clean the review comments by eliminating stop words, recognizing adjectives, and establishing semantic links. The fuzzy rule-based adaptive kernel-regularized fuzzy C-means clustering algorithm (AKRFCCA) is used to identify customer sentiments and product similarities in order to extract features and generate recommendations. In order to increase efficiency and improve forecasts for sentiment scores that could be either positive, negative, or neutral, the HQGSD-Net model combines Hamiltonian quantum generative adversarial networks (HQuGANs) with single-head self-attention (SHSA) and DOA. The dandelion optimizer algorithm (DOA) is used for the optimization process. It has been seen from the experiment that the proposed system provides accurate results in recommending customers' interests at a 99.9% accuracy level. Beyond improving sentiment-based personalization, this approach is flexible for different product categories and scalable to real-life e-commerce data.

Read at publisher →