A Novel Data Preprocessing Algorithm for Data Acquired by Vehicle Remote Monitoring System
Yuanyuan Chen · Journal of Information and Computational Science · 2014
Data is often discretized before processed by a fault prognostic model based on data-driven method. A novel discretization algorithm is proposed to work with data acquired by vehicle remote monitoring system. The novel algorithm, called NNCE, is based on normalized conditional entropy in rough set theory. Data acquired by vehicle remote monitoring system is often of high precision which results in numbers of candidate breakpoints by traditional algorithms. NNCE uses a new supervised method to reduce candidate breakpoints without changing the indiscernibility of the original data. Then normalized conditional entropy, which evaluates the distribution uncertainty of decision values, is deflned and used as the stopping criterion of NNCE. Finally, several data processing experiments are conducted to evaluate the performance of NNCE. Results show that the algorithm is efiective and has good generalization ability.