Research On Sensor Data Collaborative Prediction Technology In Resource-Constrained Scene

Internet of things(IoT)search refers to searching for resources that users need from massive IoT smart devices.In order to avoid accessing all sensor nodes,the existing IoT search architectures will establish prediction models on the cloud server,and search for sensor nodes that meet the search conditions based on the sensor data generated by the prediction model.This search framework builds prediction models one by one based on historical sensor data in the cloud.In large-scale terminal IoT systems,the model utilization efficiency is not high.Moreover,the existing IoT search architectures lack an online update mechanism for the prediction model,ignoring the problem that the highly dynamic IoT data will cause the performance of the established prediction models to gradually decline in long-term prediction.Aiming at the problem that the existing search architecture is not efficient in large-scale scenarios and cannot perform effectively prediction in a dynamic environment,the main work of this article is as follows:(1)In order to maintain the high availability of prediction models with the minimum communication cost in large-scale resource-constrained scenes,this paper designs a collaborative prediction architecture for IoT search based on the dual prediction scheme.In the architecture,prediction models established in the cloud will be delivered to the corresponding sensor.The sensor compares the difference between the predicted data and the real data every cycle to calculate the transmission value of the data,and selectively reports high-value data.The cloud uses the reported data of the sensor to update its prediction model online,and synchronizes the new model parameters to the corresponding sensor in time.This collaborative approach ensures that the cloud can always effectively predict the current reading of the sensor,and provides an accurate sensor data index for the search engine.The architecture relieves the communication pressure of resource-constrained sensors through adaptive data reporting,and the online update mechanism on the cloud server ensures that the architecture can effectively predict in a dynamic environment.(2)Based on the collaborative prediction architecture,this paper proposes a similar sensor collaborative clustering algorithm and a predictive model collaborative update algorithm:the clustering algorithm can establish a unified prediction model for sensors with similar data distribution,thereby reducing the computing pressure of the server in large-scale search scenarios;The updating method can realize the online update of the prediction model through the transfer decision based on CLCSS and various online update algorithms,and maintain the high availability of the prediction models.The simulation analysis using typical rapidly changing data sets show that compared with the existing IoT search architectures,the collaborative prediction method proposed in this paper can maintain average prediction accuracy at 94.65%and reduce communication overhead by 77%,indicating that the method can effectively reduce the network overhead while ensuring data accuracy.The designed methods in this paper can effectively reduce the frequency of sensor data reporting,ensure the high availability of the prediction model maintained on the cloud server,and enable the IoT search in a large-scale scenario to obtain search results based on the prediction data faster,saving communication overhead required by verification.