An Approximate Query Processing Algorithm with Confidence Based on Model Fitting in Sensor Networks

With the development of communication techniques, nested computation techniques and sensor techniques, wireless sensor networks have been widely applied to many fields. They can be used for testing, sensing, collecting and processing information of monitored objects and transferring the processed information to users. Collecting data of the environments is an important application of the sensor networks. Most current researches mainly focus on querying the sensing data with low energy consumption by utilizing sensor nodes' temporal-spatial correlations. These methods can collect the data with low energy consumption, but in some scenarios their results could not satisfy the applications with high confidence about the error bounds pre-specified. Moreover, these methods are not adapted to the case that there are no spatial correlations in sensor nodes. To overcome these defaults, a new method named approximate query processing algorithm with confidence based on model fitting is proposed in this paper. The proposed method create fitting models with the lower data transfer ratio, and the models are sent back to sink node instead of sensing data themselves. The proposed method can not only return the users the data within the error bounds with low energy consumption, but also be adapted to actual sensor node for being of light-weight. Theoretical analysis and experimental results show that this method can return high confident querying results and is energy efficient.