Research On Data Query Optimization Method For IoT Data Service

The Internet of Things(IoT)data can provide real-time,accurate,and comprehensive information to help people better understand and control the physical world.IoT data services are a means of rapidly acquiring IoT data,masking data heterogeneity,and providing a unified access interface for data access.Improving the response speed of IoT data queries is a challenging problem in IoT data services research.IoT data services have the characteristics of large data volume and frequent aggregation queries.Traditional accurate query methods are unable to meet users’ needs for query efficiency.Approximate query processing techniques are often applied to the analysis and processing of massive data to shorten the execution time of aggregation queries.However,existing approximate query methods still face problems such as low query accuracy and poor stability when dealing with large data aggregation queries.To solve these problems,this paper proposes an approximate query method that integrates data sampling and machine learning algorithms.The main research content is as follows:(1)To address the problems of low query accuracy and high query latency in IoT data services,a deep generative model-based approximate query method is proposed.Firstly,a query sample set is generated using a deep generative model,and then queries are performed based on the sample set to avoid obtaining samples from the data set using sampling methods,effectively reducing query latency.Experimental results show that the average relative error of the query results generated using generated samples decreases by 24.2% compared to the query results of random samples and decreases by 11.6% compared to the query results of stratified samples.(2)To solve the problem of poor stability of approximate query methods in IoT data services,a deep generative model-based aggregation query interval estimation method is proposed.Firstly,a deep generative model is trained using pre-processed data,and then multiple data samples are generated based on the model,and queries are executed on different samples.The query results are then aggregated,and the confidence interval of the corresponding query result is calculated according to the user-specified confidence level and returned to the user.Considering the problem of long processing time on a large number of samples,this paper also uses a large-scale parallel processing architecture to complete the sample generation and calculation processes.Experimental results based on a public data set show that the normalized confidence interval coverage rate of the aggregation query interval estimation results obtained using this method can reach over 85%.(3)Based on the above research results,an IoT data service system is implemented,and the design and implementation of two key modules,namely the system architecture and model initialization and the query engine,are described in detail.The actual effect of the system operation is demonstrated by simulating IoT service query requests,and the feasibility and effectiveness of the system are verified.