Abstarct: Ensemble learning is a field of machine learning that combines multiple models to improve accuracy and efficiency in predictions. Tuning appropriate hyperparameters in these models is crucial, as it directly impacts the performance of the model. In this research, the main focus is on improving the performance of ensemble learning by creating diversity in models using different hyperparameters. In this research, we propose a novel method for optimizing the hyperparameters in group learning. In ensemble learning, diversity is very important in creating a good ensemble learning model. Various factors contribute to diversity in different ensemble learning methods, such as algorithm diversity, dataset diversity, and feature diversity. It is well known that using different hyperparameters in a model can lead to diverse results. In the proposed method, this principle is used to determine the hyperparameter values. Specifically, a genetic algorithm is employed to find hyperparameters that create the highest level of diversity among models. The proposed genetic algorithm uses a fitness function that measures both diversity and accuracy. Simulated experiments demonstrate that the proposed method, by generating diversity in models through different hyperparameter settings, leads to a significant improvement in model performance. Analysis of results on 20 UCI datasets shows that, compared to other ensemble learning methods, the average f1 score of the proposed method is 2.5% better in 50% of the datasets.