The noise prediction of cassette air conditioners using AI models and analysis based on SHAP

Abstract

This paper introduces a new noise prediction method for air conditioners. In this study, XAI (eXplainable Artificial Intelligence) technology was used to identify the refrigerant flow noise of cassette air conditioners. Using the RandomForest regression method of machine learning models and DNN and CNN of deep learning models, the noise prediction models were designed, and their performance were compared. In addition, a method of analyzing factors contributing to each model was presented using SHAP (SHapley Additive exPlanation), one of XAI that can explain the predicted model. Based on these results, the idea of the analysis was suggested to find the control factors for reducing the air conditioner noise.