An energy-efficient hierarchical data fusion approach in IoT

Data Fusion (DF) involves merging data from various heterogeneous sources to generate fused data that is reduced in volume while preserving its integrity, consistency, and veracity. However, DF methodologies often pose challenges for low computational-powered sensor nodes (SNs) in energy-constrained Wireless Sensor Networks (WSNs) enabled Internet of Things (IoT). This study introduces a hierarchical data fusion (HDF) technique specifically designed to distribute the computational load among SNs with a focus on addressing the challenges of spatiotemporal data (STD). The hierarchy consists of three levels: A spatiotemporal data fusion (STDF) method, employed at the SNs level that efficiently handles the complex relationships between STD attributes; A fuzzy data fusion method, implemented at the cluster head (CH) level that effectively addresses the imprecise and fuzzy nature of real-world; The final fusion, applied at the sink (SKN) level that is based on the count of encoded icon values (EIVs). The proposed method achieves high accuracy (ACC), low error rates (ERR), and improved precision (PRE), recall (REC), and f1-score (F1S) values compared to avant-garde methods. Moreover, the analysis of the proposed technique reveals reduced computational complexity by distributing the computational load across different levels of hierarchy. Additionally, the proposed HDF technique exhibits lowered energy consumption and reduced communication overhead, making it well-suited for implementation in WSNs-enabled IoT.

Automated summary

Data Fusion is the process of merging data from different heterogeneous sources to generate fused data that is reduced in volume while maintaining its integrity, consistency, and accuracy. However, this poses challenges for low computational-powered sensor nodes in energy-constrained Wireless Sensor Networks enabled Internet of Things. This study introduces a hierarchical data fusion technique designed to distribute the computational load among sensor nodes, specifically addressing the challenges of spatiotemporal data. The proposed method achieves high accuracy, low error rates, and improved precision, recall, and f1-score values compared to avant-garde methods.