Nowadays, information overload issue is gradually exposed in the Internet of things (IoT), calling for researches on recommender system in advance for industrial IoT scenarios. With the ever-increasing prevalence of various social networks, social recommendations (SoR) will certainly become an integral application that provide more feasibly personalized information service for future IoT users. However, almost all of existing researches managed to explore and quantify correlations between user preferences and social relationships, while neglecting the correlations among item features which could further influence topologies of some social groups. To tackle with this challenge, in this paper, a deep Graph Neural Network-based Social Recommendation framework (GNN-SoR) is proposed for future IoT. First, user feature space and item feature space are abstracted as two graph networks and respectively encoded via graph neural network method. Next, two encoded spaces are embedded into two latent factors of matrix factorization to complete missing rating values in user-item rating matrix. Finally, a large amount of experiments are conducted on three real-world datasets to verify efficiency and stability of the proposed GNN-SoR.

This work proposes a novel data-driven mechanism for prediction of wastewater treatment results through mixture of two neural network models.

Optical electronic nose (e-nose) are drawing more and more attention from the academia and industry in Internet of Things applications. Since the CCD-based portable spectrometer is commonly used in gas sensing system of e-nose, the baseline drift interference is easy to be introduced, which reduces the accuracy and sensitivity of the system and makes the sensing performance worse. For this issue, firstly, the baseline drift data were analyzed and the characteristics of the baseline drift interference brought by the portable spectrometer were found. Then, according to the characteristics of the drift, a baseline drift suppression technique for portable spectrometer was proposed. Specifically, it mainly includes four steps: data preprocessing, dimension reduction using principal component analysis (PCA), sources separation using independent component analysis (ICA) and reconstruction after interference suppression. At last, experimental results show that, the proposed technique suppressed the interference effectively, and even better compared with classical existing methods.

To efficiently reduce on-grid energy consumption, the base station (BS) sleeping strategy in the hybrid energy-powered cellular network (HybE-Net) in the Internet of Things environment is studied. In order to maximize on-grid energy saving, a BS sleeping algorithm, which adaptively and cooperatively adjusts the sleeping/non-sleeping state of BSs according to the actual traffic load and solar energy state, is proposed. Specifically, the solar energy state and dynamic characteristics of the BS in solar energy supply stage are analyzed by the approximation diffusion theory. Then, by comprehensively considering communication service quality and on-grid energy saving, BS sleeping constraints are derived mathematically. Subsequently, the BS sleeping problem in HybE-Net is formulated as the on-grid energy consumption maximization problem. To solve the optimization problem, a BS sleeping algorithm based on coalitional games is proposed. Finally, simulation results demonstrate that the proposed algorithm performs excellently in improving solar energy utilization and reducing on-grid energy consumption of the HybE-Net.