Research On The Methods Of Data Mining Based On The Edge Computing For The IoT

With the advent of the Internet of Everything,the number of network edge devices has increased rapidly,causing the data generated by such devices to increase rapidly.In order to enable the Internet of Things to provide users with information services and reliable decision-making more intelligently,valuable knowledge can be fully mined from these massive data.At present,the use of cloud computing models for centralized processing of big data generated by the Internet of Things has begun to show many drawbacks.Compared with cloud computing,edge computing migrates data computing or storage to the "edge" of the network close to users,reducing the delay brought by network transmission.In addition,edge computing mode can better protect users’ privacy data and reduce the risk of sensitive data privacy leakage.Edge computing has become a new paradigm to solve the needs of Internet of things and localized computing.However,due to the limited computing resources of the edge devices of the Internet of things,the deep neural network program for data classification processing on the cloud computing platform is difficult to be directly applied in the edge devices of the Internet of things.The thesis studies the method of image data classification processing of the Internet of Things based on edge computing toalleviate the pressure of network bandwidth,and reduce the delay of data processing.The edge cloud cooperative multi-agent model is used to add early exit branches to the convolutional neural network used for image data classification;the linear regression model is used to select the appropriate exit branch points according to the network delay and equipment load,and the convolutional neural network is horizontally divided into the shallow part with input and the deep part with output through the branch points.The edge agent is established by deploying the shallow part of the neural network on the edge device,and the cloud agent is established by deploying the deep part of the neural network on the cloud server device.The multi-agent is constructed by the way of edge cloud cooperation.The reasoning execution of convolutional neural network can complete the classification and exit on the local terminal for fast local inference;when additional processing is needed,the data can be transmitted to the cloud server for further processing by using the deep neural network in the cloud to improve the performance accuracy of the system.This thesis uses the Raspberry Pi 3B as an IoT edge device to simulate IoT edge computing scenarios.Using the Chainer deep learning framework,a convolutional neural network has been built to simulate and verify the scheme proposed in the thesis,and the experiment proved the correctness and feasibility of the scheme.