Given a training dataset TRAIN_F of single-fault patterns only, a validation dataset VALID_F and a test dataset TEST_F of single- |
fault and simultaneous-fault patterns (all datasets have been preprocessed by the combination of DK and WPT and PCA, as |
presented in Figure 7(a)): |
(i) Train the probabilistic classifier |
includes pairwise classifiers as shown in Figure 6(b). |
(ii) f VALID_F, prepare the probability vector |
Calculate (f) = = (Figure 7(c)). |
(iii) For k = 1 to Run a direct search technique, such as GA or PSO, M times |
Produce an initial population for the decision threshold |
(a) , find the classification vector y(f) = y = = according to (1). |
(b) Calculate the -measure with y(f) and l(f) using (12), that is, find over VALID_F, |
where l(f) = [] is the true classification vector for input f provided from VALID_F. |
(c) Produce next generation of |
Until convergence or matching stopping criteria, return the best solution ε as the . |
(iv) Among all , k = 1 to M, choose the one producing the highest F-measure as the optimal decision threshold . |
(v) Return the trained probabilistic classifier and the optimized decision threshold as the main components of the |
intelligent diagnostic system. |
(vi) The performance of and can be evaluated with TEST_F and as illustrated in Figure 7(d). |