Tree-Based Data Aggregation Approach in Periodic Sensor Networks Using Correlation Matrix and Polynomial Regression

Sensor networks are a collection of sensor nodes that co-operatively transmit sensed data to a base station. One of the well-known characteristics of Wireless Sensor Networks (WSNs) is its limited resources. Energy consumption of the network's nodes is considered one of the major challenges faced by researchers nowadays. On the other hand, data aggregation helps in reducing the redundant data transferred through the WSNs. This fact implies that aggregation of data is considered a very crucial technique for reducing the energy consumption across the WSN. Local aggregation and Prefix filtering are two methods used in which they utilize a tree based bi-level periodic data aggregation approach implemented on the source node and on the aggregator levels. In this paper an efficient model for multivariate data reduction is proposed based on periodic data aggregation on two sensor levels, in addition to polynomial regression functions. The performance of the model was evaluated using SensorScope network which is deployed at the Grand-St-Bernard located between Switzerland and Italy. The results show the advantages of the proposed model as it allows 84% reduction rate and 93% approximation accuracy after reduction. The simulations were done using the R software.