Data Congestion Prediction in Sensors Based IoT Network

Abstract

Internet of Things (IoT) becoming the major part of human life and make the life simpler. IoT uses sensor nodes to monitors certain phenomena and transmitting the collected information to the IoT gateway. The size of the network is increase rapidly and causing congestion in the network and results in network delay, loss of data packets, a decrease in throughput, and poor energy efficiency. It is important to predict the congestion and mitigate the data accordingly. To resolve the problem of congestion, our focus is on predicting the congested node effectively. We propose an optimized deep neural network - Restricted Boltzmann machine (DNN-RBM) based data congestion prediction approach which is used for analyzing and predicting the congested node in the sensors based IoT environment. To enhance the performance of DNN, the weight parameters of DNN are optimized using the Restricted Boltzmann Machine (RBM)-algorithm. The dataset is used to train the model and enable the prediction to find the congested nodes in the network with more accuracy to enhance the performance of the network. The performance factors congestion window, throughput, propagation delay, RTT, number of packets sent, and packet loss are given as input by using DNN-RBM. Predicted results show that the proposed DNN-RBM model predicts congestion with more than 95% accuracy as compared with other models like ANN, DNN-GA.