Configuration of Neural Network Hyperparameter using Ant Colony Optimization Algorithm

Deep neural networks are widely used in classification and prediction tasks. These problems are typically solved using multilayer neural networks. However, one of the most challenging tasks is setting the hyperparameters of these networks. This process is time-consuming because hyperparameters are still often chosen through a trial-and-error approach. In this paper, we propose a metaheuristic optimization technique to automatically set the hyperparameters of a neural network for prediction tasks. Specifically, we introduce a new ant colony optimization algorithm (ACO-NN) to configure the number of neurons in the hidden layers of the network. The proposed approach is validated using data from undergraduate students, which includes demographic, socioeconomic, microeconomic information, data at enrollment, and academic records from their first and second semesters. We evaluate the performance of our method on this dataset and compare our results with those of other machine learning algorithms. In all cases, our approach achieves better results in terms of test accuracy, precision, recall, and F1-score.