[2P6-GS-13] AI application: Real estate

[2P6-GS-13-01] Estimation of building structure and age using real estate geospatial information and building images

〇Yoshiki Ogawa¹, Takuya Oki², Yoshihide Sekimoto¹, Ryosuke Shibasaki¹ (1. the University of Tokyo , 2. Tokyo Institute of Technology)

Keywords:GIS, Realestate point data, Building damage, Sparse modeling, Building usage

In this paper, we propose a method for estimating the structure and age of buildings on a building unit using machine learning from Geographic Information System (GIS) data and image data obtained from real estate data. Specifically, we proposed two models. Model1: the real estate point data (with structure and age) is spatially joined with GIS data, such as restricted areas and roads, and information in real estate data, to create training data and applying Sparse modeling (SpM). Model 2: a model that estimates the building structure and age of each building is developed by Convolutional neural network (CNN) from the image data set from real estate data. Finally, the accuracy of the developed model is verified check by demonstrating that the type of building structure and age can be classified into each type with accuracy by applying the proposed identification method.