A0250
Title: Comparison of prediction methods for spatial data using real estate data
Authors: Koki Kirishima - Hiroshima University (Japan) [presenting]
Mineaki Ohishi - Tohoku University (Japan)
Hirokazu Yanagihara - Hiroshima University (Japan)
Abstract: Data with location information such as latitude and longitude are called spatial data. We consider a prediction problem for real estate value which is a typical example of spatial data. As a method of spatial data analysis, Geographically Weighted Regression and Spatially Clustered Regression have been proposed. In addition, an estimation method using the adjacency relations in the space by Generalized Group Fused LASSO has also been proposed. The second and third methods can also be used to cluster spatial data. By applying these methods to real estate data, we compare the accuracy of predictions based on each method. In addition, we also compare prediction methods based on machine learning, such as random forests and neural networks, which have high prediction accuracy.
