Özet:

In this study, the artificial nerve network models and the one-input and double-input volume estimates according to the length of the tree along with the diameter of the trees, the effects of different transfer function and the estimated success of the number of neurons were studied. For this purpose, data training for 500 different YSA models alternatives, including five different transfer functions and 100 different neuron number alternatives, was carried out and forecasts were obtained. In comparison based on the various success criteria, whether single-input or double-input volume estimates have provided the most successful estimates of the transfer function in which the linear (pure-lin) is at the input layer and the intermediate layer connection point and the hyperbolic tanjant sigmoid (tan-sig) is at the intermediate layer and the output layer connection point. On the other hand, with the increase in the number of neurons, a remarkable volume forecast success improvement was not achieved. The HKT value for the most successful single-input YSA model; 3.0084, RMSE value; 0.1262, R2 straight. The value; 0.9720, the AIC value; -785.6687 and the SBC value; -777.1746, the THY value; 0.0889 and the OMHY value are calculated as 9.2617. The HKT value for the most successful double-input YSA model; 1.4510, RMSE value; 0.0879, R2 straight. The value; 0.9864, the AIC value; -924.2006 and the SBC value; -905.2125, the THY value; -0.51682 and the OMHY value are calculated by 5.8764%. As a result of the success of the estimates that will be obtained by the YSA models of the transfer function and the number of neurons, which is the main subject of this study, it is also concluded that instead of increasing the number of neurons, the choice of the appropriate transfer function that allows the transition between the input, secret and output layers is important.

Anahtar Kelimeler: