Sahar Askari - Department of Mathematics, Farhangian University (Shahid Sadougi Pardise) of Kermanshah Branch, Kermanshah, Iran,
Nina Daryakenari - Department of Mathematics of faculty of basic sciences South Tehran Branch, Islamic Azad University, Tehran, Iran,
Parvin Azhdari - Department of statistics, Tehran North Branch, Islamic Azad University, Tehran, Iran.
Somayeh Ezadi - Department of statistics, Tehran North Branch, Islamic Azad University, Tehran, Iran.

In this research, first by using Z-number a generalized neural network is proposed. This generalized neural network is a two layer neural network. First layer consists of input and the second layer includes output, including linear activation function, weights and output. Inputs are real valued but weights and outputs are Z-number. In this case, to estimate simple linear regression coefficient, first the equation is considered as four real equations then only with an objective function, the network is ready for training process of least squares error method. The present study improved the method which was put forward by Ezadi and Allahviranloo. Approximate solution by in this study is compared with their approach (the only present method). The finding revealed the proposed method by this study can be beneficial in terms of time and computational cost

Fuzzy-numbers, Z-numbers, Linear Regression, Neural Network