A0338
Title: Applicability of TreeSHAP to analyze real estate data
Authors: Koki Kirishima - Hiroshima University (Japan) [presenting]
Mineaki Ohishi - Tohoku University (Japan)
Ryoya Oda - Hiroshima University (Japan)
Kensuke Okamura - Tokyo Kantei Co Ltd (Japan)
Yoshimichi Itoh - Tokyo Kantei Co Ltd (Japan)
Hirokazu Yanagihara - Hiroshima University (Japan)
Abstract: Machine learning models such as random forests are often used to analyze real estate data. Although machine learning models achieve very high predictive accuracy, the difficulty lines interpreting the predictions. The SHAP, which is one of the methods of XAI, has been proposed to overcome this difficulty and facilitates the interpretation of the predictions by linearly decomposing the influence of each explanatory variable on the predictions. For random forests, TreeSHAP, which allows SHAP to be computed strictly and quickly, has been proposed. The applicability of TreeSHAP to analyze real estate data is discussed.
