Abstract:

The core problem in using fuzzy integral is to determine the fuzzy densities. Through analyzing the disadvantage of the foregone methods of determining fuzzy densities, a new method of determining fuzzy densities adaptively is presented depending upon evidences support degree which is defined based on the interrelation of evidences. The influence of diversity measure among sorters on the evidence support degree is analyzed, and an influencing factor is put forward to perfect the meaning of the evidence support degree. Experimental results indicate that the proposed method has high precision compared with the single support vector  machine, voting-based support vector machines, support vector machines static fuzzy integral using the precision of classifiers as fuzzy densities, and fuzzy integral based on fuzzy densities determined adaptively using for support vector machines ensemble.