Abstarct: In this thesis, a hybrid scheme baxsed on the recurrent neural networks for approximate fuzzy coefficients (parameters) of fuzzy linear and polynomial regression models is presented. When multicollinearity happens (As in regression models in traditional statistical theory), the model may lead to poor prediction. To overcome the problem, Bridge regression methodology which combines Bridge regression with the fuzzy regression models in order to reduce the effect of multicollinearity, for the first time is considered in this thesis. Then, the proposed model and its specific formulations (Ridge and Lasso regression) are solved by using a capable neural network. The proposed neural network is first constructed baxsed on some concepts of convex optimization and stability theory. The presented neural network frxamework guarantees to find the approximate parameters of the fuzzy regression models. The existence and convergence of the trajectories of the neural network are studied. The Lyapunov stability for the neural network is also shown.