Time-Aware and Location-Based Personalized Collaborative Recommendation for IoT Services

IoT applications need to actively monitor and respond to service invokes to guarantee the reliable connectivity of data and devices. However, with the gradual increasing of dataset, it is difficult to provide accurate and effective service in time. In order to solve the problem of information overload, the recommendation system has been proposed. In recent years, there are some progresses in the research of recommendation system based on collaborative filtering (CF), but most of them face sparse problems and scalable problems. In this paper, a personalized recommendation model is given building on location and time information. Data sparsity is alleviated by padding the missing value of the user-service-time tensor over the adjacent time period. Users set and services set are divided into multiple clusters respectively based on location information and similar items are selected in smaller and highly similar clusters, which makes our model scalable. The time decay function and the personalized weight is exploited in our method to improve the prediction accuracy. Massive experiments based on real-world data indicate that our method can effectively improve the accuracy of prediction compared to other models.