With the continuing advancements in technologies (such as machine to machine, wireless telecommunications, artificial intelligence, and big data analysis), the Internet of Things (IoT) aims to connect everything for information sharing and intelligent decision-making. Swarm intelligence (SI) provides the possibility of SI behavior through collaboration in individuals that have limited or no intelligence. Its potential parallelism and distribution characteristics can be used to realize global optimization and solve nonlinear complex problems. This paper reviews representative SI algorithms and summarizes their applications in the IoT. The main focus consists in the analysis of SI-enabled applications to wireless sensor network (WSN) and discussion of related research problems in the WSN. Also, we concluded SI-based applications in other IoT fields, such as SI in UAV-aided wireless network. Finally, possible research prospects and future trends are drawn.

In resource-limited environment, grid users compete for limited resources, and how to guarantee tasks’ victorious probabilities is one of the most primary issues that a resource scheduling model cares. In order to guarantee higher task’s victorious probabilities in grid resources scheduling situations, a novel model, namely ESPSA (Extended Second Price Sealed Auction), is proposed. The ESPSA model introduces an analyst entity, and designs analyst’s prediction algorithm based on Hidden Markov Model (HMM). In ESPSA model, grid resources are sold through second price sealed auction. Moreover, to achieve high victorious probabilities, the user brokers who are qualified to participate in the auctions will predict other players’ bids and then carry out the most beneficial bids. The ESPSA model is simulated based on GridSim toolkit. Simulation results show that the ESPSA model assures a higher victorious probability and superior to other traditional algorithms. Moreover, we analyze the existence of Nash equilibrium based on simulation results, thus, any participant who changes its strategy unilaterally could not make the results better.

Predicting customer Churn is of principal importance for many businesses, as it directly influences the continued utilization of company products by consumers and has a significant impact on company’s financial health and sustainability. However, accurately forecasting customer attrition within the telecommunications sector presents considerable challenges. In response to this need, we propose a novel and highly accurate churn prediction model utilizing a Multilayer Perceptron (MLP) architecture. Our methodology begins with thorough data preprocessing, followed by the application of MLP to predict customer Churn. By employing the Synthetic Minority Over-sampling Technique (SMOTE) and iterative balancing, our proposed model has demonstrated an accuracy improvement of at least 5%, alongside a minimum 5% increase in the F1-score, when compared to existing state-of-the-art algorithms on the widely recognized Telco-Customer-Churn dataset. This research not only expands the opportunities for enhancing the interpretability of predictive models but also fosters a deeper understanding of Churn prediction mechanisms. Ultimately, these advancements contribute to the commercial sustainability and financial stability of telecommunications companies by facilitating more informed decision-making processes regarding customer retention strategies.

Sensors in the Internet of Things (IoT) generate large amounts of data, which requires high-speed data transmission. In order to achieve the parallel transmissions of the wireless sensor network on the transmission layer, the performance of stream control transmission protocol (SCTP) and transmission control protocol (TCP) in the wireless sensor network under different packet error rates was simulated and compared. A dynamic multipath handover method for SCTP (MS-SCTP) was proposed to improve the transmission performance, which selects the transmission path according to the packet error rate and the retransmission ratio in the sender’s buffer. The TCP and SCTP protocol switching method (TCP-SCTP) was proposed to detect the current network traffic and adjust the MS-SCTP or TCP method. Analysis and simulation results show that MS-SCTP and TCP-SCTP could improve network throughput and reduce packet loss rate. MS-SCTP and TCP-SCTP can be combined with other technologies and channel allocation algorithms to improve network traffic.