Evaluation of Lot Shape for Automated System of Land Appraisal with Neural Networks

点击翻译按钮获取中文摘要

The aim of this research is to develop a method to evaluate the shape of lots in appraisal process: developing appraisal function using such as neural networks, analyzing the shape evaluation mechanism, and developing a partial system of land appraisal. This year as the final year of the project I refined evaluation function of land price using detailed database of land prices in Setagaya-ku in Tokyo, through which two major results have been obtained. One set of results is to estimate a very precise hedonic land price model of detailed data on lot shape, locational factors and environmental factors. Significance of the model lies in its incorporation of micro-level residential environment that enabled us to evaluate in monetary terms the external effects of subdivision of land parcels as well as allocation of parks. In particular, the effects of greenery and lend-use mixture have different influence depending on the size of the lot. For larger lots, the greenery effect is smaller, and the land-use mixture effect is larger. This model can be used to city planning/right adjustment supporting system in district planning, and can be applied to obtain optimal block division problems. The other set of results concerns the appropriate choice of reference lots in appraisal process. It was empirically shown that appropriate choice of reference lots enables us to evaluate the land price more precisely than a hedonic regression. This result justifies the sales comparative method in appraisal process. Moreover, the results revealed what kind of factors of reference lots should be conformed to the lot in evaluation: Distance to the nearest railway station, time to the central city, and shape of lots such as corner lot and flag shape should be equalized as much as possible for the precise appraisal.