Özet:

In this study, a data set containing air pollution information collected from different locations was used. This dataset contains air pollution measurement information and results information for 25 different districts of Seoul city. The values obtained from similar regions were filtered from the entire data set and a data set specific to these regions was created. As the next step, the model was trained with the values obtained in its own regions and the model was recorded. When this model is tested with the data obtained in its own fields, it shows very successful results. However, when tested with data obtained from different regions and dissimilar to these regions, the model classification success remained very low. The aim of the study is to show that more successful results can be obtained with the distributed learning method, which can be a solution to this problem situation, compared to classical machine learning. In addition, security problems during model aggregation and aggregation should not be ignored in the Distributed Learning section. For this reason, a robust blockchain-based approach is proposed against malicious attacks during model deployment.

Anahtar Kelimeler: