H.3.13. Intelligent Web Services and Semantic Web
Atefeh Niroomand; Seyyed Hamid Ghafouri; Amid Khatibi Bardsiri
Abstract
This study addresses the challenges of managing dynamic and heterogeneous Internet of Things (IoT) data by proposing a time-aware recommender system that integrates a dynamic semantic ontology with clustering techniques and a hybrid collaborative filtering framework. The proposed model continuously updates ...
Read More
This study addresses the challenges of managing dynamic and heterogeneous Internet of Things (IoT) data by proposing a time-aware recommender system that integrates a dynamic semantic ontology with clustering techniques and a hybrid collaborative filtering framework. The proposed model continuously updates the ontology based on user interactions and incorporates temporal information into both knowledge representation and clustering processes, enabling adaptive and real-time modeling of evolving user behaviors.The dataset consists of approximately 500 users and 15,000 time-stamped interaction records collected over four months, including demographic attributes (age and gender), IoT device usage patterns, and temporal features such as timestamp and time of day.The recommendation framework combines ontology-enhanced user-based collaborative filtering with dynamic K-means clustering, leveraging both semantic relationships and behavioral similarities to improve recommendation quality. Experimental evaluation is conducted using Precision, Recall, F1-score, Accuracy, MAE, and RMSE metrics. The model achieves improvements ranging from approximately 2% to 52%, with respect to state-of-the-art non-temporal methods and traditional collaborative filtering techniques, respectively.Furthermore, computational complexity analysis indicates that the additional processing cost introduced by dynamic ontology updates and temporal modeling remains manageable, preserving the practical applicability of the proposed framework in resource-constrained IoT environments.
