The Advantages of Artifical Neural Networks to Give Length-Weight Relations and Comparison of Growth

Abstract

This study has two main purposes. The first one is to compare the performances of Logistics, Gompertz, Schnute and von Bertalanfy models according to the determination coefficient (R^2) and Mean Squared Error criteria, and these comparisons are aimed to be discussed in terms of gender.
For this purpose 329 female and 319 male individuals’ data on P. Quadrilineatus (Bloch, 1790) which is one of lesepsian species in Turkish marine waters were used. As a result of the comparisons, it was concluded that the best growth model varied on the basis of genders. For example, for male individuals Gompertz is the best growth model with highest R^2 (0,745) and the lowest Mean Squared Error (0,717) values. For females, Schnute and von Bertalanffy models have the best results with the highest R^2 (0.813) and the lowest Mean Squared Error (0.369) values.
The second purpose of this study is to make comparisons regarding the use of Artificial Neural Networks and Linear Regression Method to determine the Length -Weight relationships. Comparisons are made according to R^2, Mean Squared Error and Mean Absolute Percentage Error criteria. Although the results of these criteria are very close to each other, it can be said that the results obtained for Artificial Neural Networks are slightly better for overall data. The use of Artificial Neural Networks is more convenient since it is a faster and less annoying (does not need transformation) method compared to Linear Regression Method, especially in length-weight estimation for Pelates quadrilineatus (Bloch, 1790) species.