| Title |
Typification of Irregular Shaped Land Parcels Using Machine Learning |
| Authors |
홍성조(Hong, Sungjo) ; 이윤서(Lee, Yoonseo) |
| DOI |
https://doi.org/10.5659/JAIK.2022.38.3.189 |
| Keywords |
Irregular Shaped Land Parcel; Shape of Land Parcel; Machine Learning; K-means Clustering; Urban Spatial Structure |
| Abstract |
The shape of land parcels greatly affects land value, the density of buildings, and the shape of a building. Although the Korean system
classifies parcel shapes into 6 types, there are irregularly shaped land parcels that cannot be classified. Irregular shaped land parcels impose
many restrictions on the arrangement and form of buildings, and these restrictions are even more severe with small parcels. Until now,
studies on the shape of parcels have been conducted, but studies on irregularly shaped land parcels have been insufficient. Therefore, this
study aims to typify irregular shaped land parcels that are difficult for humans to distinguish by applying machine learning methodology and
to identify the characteristics of each type. The subject of this study is irregular shaped land parcels in the class-II general residential areas
of Seoul; there were 500 sample parcels extracted and used for analysis. Irregular shaped land parcels were typified using K-means
clustering, which is a representative method of unsupervised learning to solve classification problems. Afterwards, the values of Shape Index
(SI), STandard Index (STI), and With-depth Ratio (WR), which are indices related to parcel shape, were compared by type. Upon analysis,
the types of irregular parcels could be divided into avocado type, potato type, corner type, bell type, stick type, and L-shaped type. The stick
type and L-shaped type reflected small SI values. The avocado type, corner type, and L-shaped type revealed small STI values. Lastly, the
WR value was substantial for the stick type and L-shaped type. |