TY - EJOU AU - Somayeh Ezadi, AU - Tofigh Allahviranloo, TI - Numerical Solution of Linear Regression Based on Z-Numbers by Improved Neural Network T2 - Intelligent Automation \& Soft Computing PY - 2018 VL - 24 IS - 1 SN - 2326-005X AB - In this article, the researcher at first focuses on introducing a linear regression based on the Z-number. In this regression, observations are real, but the coefficients and results of observations are unknown and in the form of Z-rating. Therefore, to estimate this type of regression, we have three distinct ways depending on different conditions dominating the problem. The three methods are a combination of artificial neural networks and fuzzy generalized improvements of the technique. Moreover the method of calculating the weights of the Z-number neural network has been mentioned and the stability of neural network weights is considered. In some examples, the answer is estimated compared with the original answer. KW - Z-numbers; linear regression; neural network; error analysis; weights; stability DO - 10.1080/10798587.2017.1328812