In this paper, random forests are proposed for operating devices diagnosticsin the presence of a variable number of features. In various contexts, likelarge or difficult-to-access monitored areas, wired sensor networks providingfeatures to achieve diagnostics are either very costly to use or totallyimpossible to spread out. Using a wireless sensor network can solve thisproblem, but this latter is more subjected to flaws. Furthermore, the networks'topology often changes, leading to a variability in quality of coverage in thetargeted area. Diagnostics at the sink level must take into consideration thatboth the number and the quality of the provided features are not constant, andthat some politics like scheduling or data aggregation may be developed acrossthe network. The aim of this article is ($1$) to show that random forests arerelevant in this context, due to their flexibility and robustness, and ($2$) toprovide first examples of use of this method for diagnostics based on dataprovided by a wireless sensor network.
展开▼
