Federated learning enables multiple clients to collaboratively train a machine learning model while keeping their local data private-omitting sharing data, making it a privacy-preserving technique. Data plays a key role in these models. However, in some cases, a single organization may not have enough data or high-quality data to build a reliable model, especially in a rapidly changing environment. In horizontal federated learning, each organization/client continuously refines its model, which is periodically fused and distributed among all participating clients in the federation for further enhancement. The fusion/aggregation process typically relies on a weighted averaging approach, where the weights are determined by the quality of each client’s model. This study explores approaches by using federated learning with respect to uncertainty and examines various aggregation strategies based on the performance of local models.