Towards Utilization of Neurofuzzy Systems for Taxonomic Identification Using Psittacines as a Case Study

Table 4

Round 4. Data separated by gender for some species (see the Appendix); 3,660 patterns presented for use in training; NEFCLASS-J default for learning rate was kept; number of MF for hue variables was 16; MF shape was trapezoidal; aggregation function was maximum; learning procedure was best per class; all four learning constraints were used; maximum number of epochs was 500, minimum number of epochs was 10, and number of epochs after optimum was 25.

Run	1	2	3	4	5	6
Max. # fuzzy rules	Auto.	1500	1000	500	250	183
Correctly class.	97.32%	81.23%	65.25%	41.61%	33.06%	20.87%
Num. rules	1,917	1,499	1,051	500	362	183
Mean error	0.358197	0.427322	0.676230	0.766667	0.829508	0.882240