Abstract

Recently, with the development of artificial intelligence architecture and the rapid improvement of hardware performance, various artificial intelligence models are attracting attention and their usefulness are increasing. However, unlike the rate of development of performance, the development of model's explanatory ability is slow to progress. The increasingly complex decision branching points of AI models are increasing exponentially, which hinders researchers' ease of model interpretation. Explainable Artificial Intelligence, XAI has emerged to solve the above complexity, and helps to improve interpretability and reliability by decomposing the black box of the model to a level that researchers can understand. This study estimates the surface temperature of a specific spatial unit using land cover data of a cell unit in Korea, Seoul. The limitations of the existing statistical model for high-dimensional data were examined, and the estimation results were compared using a machine learning model. Models used include Lasso regression(Least absolute linkage and selection operator), random forest regression, and XGBoost (eXtreme Gradient Boosting) regression. Finally, based on the fitted result of the XGBoost regression model, XAI SHAP was carried out.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call