Modeling and Forecasting Eggplant Production Quantity in Türkiye with Artificial Neural Networks

Abstract

This study aimed to investigate eggplant production in Turkey using artificial neural networks. In this context, artificial neural networks (ANN) were applied to the dataset consisting of 64 the number of observations covering the period 1961-2024 using the hyperbolic tangent activation function. Here, years were treated as the independent variable, and eggplant production amount as the dependent variable. The ANN model was a network architecture with a single hidden layer, six processing elements (1-6-1), and the Levenberg–Marquardt backpropagation algorithm (trainlm) was used as the learning algorithm. Model performance was measured using Mean Square Error (MSE) and Mean Absolute Error (MAE). According to the results, the optimal MSE (1 196 450 266.58) and MAE (26 977.843) values were achieved. Forecasts made with the hyperbolic tangent function indicate an increasing trend in eggplant production during the 2025-2030 period. The average annual increase for this period is expected to be between 2600-3000 tons. These results demonstrate that artificial neural networks provide good results in modeling and forecasting crop production amounts.