Abstarct: In recent years, decentralized architecture has attracted a lot of attention. Web design and machine learning design areas can be considered as areas where this architecture has had significant development. The Federated module in Webpack Version 5, Astro, and Fresh are among the frxameworks developed in the field of web-baxsed content design. In the field of machine learning, Federated learning that uses decentralized education can be mentioned. Federated learning is one of the emerging methods that, while maintaining data security, also reduce costs using client resources in education. Research using this method in categorizing such as fraud detection and communication demonstrates the efficiency of this method. In a practical environment, one of the challenges that can be made for clients is their adjustment. The isolation and variety of data in clients are one of the most critical obstacles to client adjustment. In this research, we try to automatically adjust the clients by applying the genetic algorithm. This study's results show that client parameter adjustment improves Federated learning accuracy at the client and server levels.