Abstract: The monitoring system by using sensor networks belongs to large scaled complex network, where sensing data are delivered at a fixed time interval in a predefined way to the sink node for user queries, thereby the application environment can be monitored. Due to the fact that the quality of service is affected by sensor networks and the application environment, the sensor data collected is usually with uncertainty. In addition, the mechanism of periodically report may also lead to inaccurate information for real time application. In this paper, we apply time series model to forecast the sensor data, then response the user queries, which will reduce the communication overhead. By analyzing sensing data produced by sensor networks, a multi-attribute fuzzy time series forecasting model is also introduced, and it takes the trend factor existing in time series into consideration. An improved model that suits the forecast of sensing data is ultimately proposed. Simulation results show that the proposed fuzzy time series forecasting model can effectively predict future sensing data of sensor networks and improve the predicting accuracy.

Key words: information management and information system, fuzzy time series forecasting model, sensor networks, distributed database

摘要： 传感器网络监控系统属于大型复杂系统,由感知节点以一定的时间间隔向sink节点发送感知数据,以实现对应用环境的监控。由于网络本身及应用环境的影响,得到的感知数据往往存在不确定性。此外,周期性报告数据模式影响到实时监控数据的精确性。本文应用时间序列模型预测传感器数据以响应用户查询,可有效降低网络通信量。通过对无线传感器网络的数据分析,引入多属性模糊时间序列预测模型,充分考虑了无线传感器网络时间序列中存在的趋势因素,并提出了适合于传感器网络的修正预测模型。实验结果表明模糊时间序列模型可有效预测传感器网络数据,且能提高预测精度。

关键词: 信息管理与信息系统, 模糊时间序列预测模型, 传感器网络, 分布式数据库