Comparing Logistic Growth Model and Competing Models for Predicting the Overall Index in the Tehran Stock Exchange

The main objective of this study is to compare the Harvey Logistic Growth Models, Harvey, and the Nonlinear Autoregressive Neural Network, and to design and find a model with better predictive accuracy for the Tehran Stock Exchange data. This model is nevertheless highly dependent on past values, has high fluctuations, and shows nonlinear motion patterns which have been repeatedly neglected.

In this study, the "Harvey Logistic" Growth Models, Harvey and the addition of nonlinear components based on the Taylor series expansion for trigonometric functions were studied to compare the accuracy and prediction of these models based on prediction criteria and its results with the nonlinear autoregressive neuronal network. Daily data of fluctuations from 1393 until 1395 of the total stock index, which was divided into two categories, were used as the sample pool in this study.

The results of the unit root tests such as Dickey-Fuller and BDS test show that the data is stationary and has a nonlinear property. In the estimation stage, the goodness of fit for the Logistic and Harvey models show that both models have a high root mean square error and low coefficient of determination for the four data types. By adding the nonlinear parts to the Harvey model, a good fit was obtained for the stock index with a coefficient of determination of about 99.8 percent and minimum root mean square error, even when compared with the nonlinear autoregressive neural network.

The results of the research show that combining the Harvey model with the nonlinear component could be considered as one of the models which predict the Tehran Stock Exchange index better than the other models.