JoWUA

Abstract

The use of smart internet services to deliver customized information and enhance user interaction is also gaining full relevance in e-commerce sites. Although a lot has been achieved in recommendation systems, current solutions are largely designed as standalone prediction models and do not address critical service-level issues, such as dynamic user behaviour adaptation, online, scalable deployment, and efficient integration with internet-based e-commerce platforms. This shortcoming has shown an essential disjunction between the accuracy of the recommendation and the viable service-based implementation. To fill this gap, this paper introduces a new internet service-oriented recommendation framework tailored to e-commerce platforms. The proposed system will be an online service that integrates user behaviour modelling and adaptive learning to produce recommendations that are correct and context-sensitive. In contrast to traditional methods, the new approach focuses on service scalability and real-time adaptability, enabling flawless implementation in web-based e-commerce systems. The workflow is a mathematical model and algorithmic process that defines the recommendation process to be robust and understandable. The unique feature of the suggested approach is that it follows a service-based design, turning the recommendation mechanism into a scalable internet service rather than a fixed analytical model. Such a design enables effective management of dynamic user interactions and enhances service quality in e-commerce applications. Benchmark experiments on new recommendation methods show that the suggested framework outperforms current methods in terms of accuracy, stability, and efficient service performance. The findings demonstrate the efficiency of the proposed system in reducing the gap between recommendation intelligence and practical implementation of internet services. DCRN improves HR@5, HR@10, and NDCG by 7.91, 6.16, and 8.55, respectively, on the WeChat Channels dataset. The gains are 6.11% in HR@5, 6.08% in HR@10, 4.29% in NDCG@5, and 3.99% in NDCG@10 in the Tmall data. Equally, for the CIKM data, the Proposed System outperforms the Existing Model, achieving gains of 6.65% in HR@5, 5.57% in HR@10, 7.34% in NDCG5, and 6.62% in NDCG10.