The Japanese government publishes official land prices each year via the Land Market Value Publication. However, the officially published prices cover only a small number of properties. In pricing a property, real estate companies thus refer to both the official land prices of comparable properties and the transaction history of the property in question, as buyers face the problem of not knowing the standard price of the land. Therefore, if we buy a piece of land, we need to estimate its price based on the published official land prices. In this study, we estimate land price by kriging using road network distance and verify the precision of the estimation.

Kriging is a linear regression method applied to space. The method models the temporal and spatial relevance of natural phenomena. Kriging is a major technique in Geo-statistics that employs the covariance function, and is mainly used to estimate ore reserves. Generally, spatial properties depend on direct distance. Therefore, estimation by kriging depends on distance. In this paper we compare the estimation accuracy of Euclidean distance and road network distance.

In conclusion, the numerical experiment using road network distance clearly achieved a better estimated result  than using Euclidean distance in urban space.