Abstarct: Today, due to high competition in global markets and technological innovation, the issue of measuring the Performance of organizations has become a global issue. One of the most common methods of measuring performance is data envelopment analysis. But it has problems such as sensitivity to the border of performance and lack of predictive power, which one of the ways to solve these problems is to combine it with an artificial neural network. Border neural networks have more robust and flexible performance than data envelopment analysis and it can also be used to solve larger scale problems The purpose of this study is to evaluate the performance of companies listed on the stock exchange using a neural network. In this research, for the given time series, after preparing the data, the optimal values of the phase space parameters, including time delay, are obtained. Then, by creating an input matrix of time series data and a suitable structure for the forecasting network, modeling and forecasting time series should be done. In this regard, two proposed models baxsed on nonlinear methods for predicting time series are presented. The first model of nonlinear autoregressive neural network is compatible with exogenous inputs and the second model of neural networks is adaptive neuro-fuzzy inference system. In order to have a criterion for comparison, we used the mean squared error. Finally, reviewing and evaluating the results shows the appropriate structure of the models used and the use of fuzzy space in increasing the accuracy of modeling and predicting time series is effective and at the same time, the performance of the neural-fuzzy network model is significantly better than the time delay neural model. The results indicate that time latency and neural-fuzzy neural networks can be used as reliable modeling tools to predict performance.