In this paper we develop a model that predicts the shape of detached houses
from the shape of residential lots by introducing the framework of the ass
ociative memory model. The model is based on the idea that, if the shapes o
f lots are similar, then the positions of associated houses may also be sim
ilar. To this effect, 'similarity indices' of lots are devised and transfor
med into probabilities. The overlay of sample house positions weighted by t
he probabilities predicts the probability density for each point in the lot
covered by a house. The model is tested empirically with data from Hiratsu
ka city in the Kanagawa prefecture in Japan. The results of its sensitivity
analysis by gradually changing the lot shape reproduces the actual tendenc
ies of allocation of houses and yards for lots of limited area.