Accuracy-enhanced E-commerce recommendation based on deep learning and locality-sensitive hashing

Recommender systems facilitate the discovery of relevant content in several online communities by analyzing users' past interactions and preferences. With the expansion of data-intensive online activities and online content, cybersecurity risks have increased. Users may not be adequately protected by traditional collaborative recommendation systems. Sparsity and cold-start are common challenges for traditional recommendation systems. Advances in deep learning have enabled recommender systems to enhance user behavior prediction precision, a task previously deemed unattainable. To enhance privacy and speed up neighbor searches, we propose locality sensitive hashing (LSH) in neighbor-based embedded learning. Through an adversarial approach, LSH enables efficient neighbor searching. Deriving multi-view embeddings from diverse behavioral data enhances the accuracy of predictions. By using multi-view preference embeddings, user preferences can be depicted more intricately. LSH, neighbor-centered embedding, self-embedding, and interaction-aware embedding are all used to accomplish this task. In addition to providing efficient similarity search capabilities, neighbor-based embedding learning and adversarial search provide robust privacy protection. As a result, the outcomes are consolidated into an advanced prediction system based on long short-term memory. Numerous empirical studies with authentic datasets demonstrate that our proposed methodology outperforms existing state-of-the-art benchmarks in terms of predictive accuracy, while maintaining robust security.