An improved self-organizing CPN-based fuzzy system is proposed in this paper.Asso-ciated with the neuro-fuzzy system,there is a two-phase hybrid leaing algorithm,which utilizes aCPN-based nearest-neighborhood clustering schemefor both structure leaing and initial parameters set-ting,and a gradient descent method with variableleaing rate for parameters fine-tuning.By combin-ing the above two methods,the leaing speed is muchfaster than that of the original back-propagation al-gorithms.The comparative results on the examplessuggested that the method has the merits of simplestructure,fast leaing speed and good modeling ac-curacy.