EasyRBF: Towards Infilling Missing Soil Data

Soil characteristics are one of the most important evidential attributes for hydrologists and other environmental scientists to conduct domain-specific studies. To keep soil data complete is the first step towards analyzing soil characteristics. Usually collected by wireless sensor networks, however, the soil datasets often inevitably suffer significant and continuous data missing due to the unreliability of wireless channels or the failures of sensing instruments. In this paper, we present a novel missing value infilling approach for soil datasets, called EasyRBF, which finally relies on a RBF network to predict or estimate the missing soil data. The key ideas of EasyRBF are (1) separating the optimizations of RBF network parameters in stages, and (2) leveraging two delicately nested PSO (particle swarm optimization) procedures to simultaneously solve the critical RBF network parameters, different than traditional PSO-based learning schemes. We conduct extensive numeric experiments over a real-world soil dataset, to evaluate the performance of EasyRBF and compare it with other five typical approaches for soil data infilling. The results demonstrate that EasyRBF can achieve higher accuracy of infilling the soil dataset with large-scale and continuous data missing.