필지의 기하학적 특성 정보와 의사결정트리 기법을 활용한 토지 형상 인식 방법론 연구

Abstract

The land feature survey needs to be conducted in an objective and consistent manner. Among the categories of characteristics survey, however, the land shape has had vague reference points, undermining the objectivity and consistency of the survey. As such, this study intended to extract the geometric characteristics information of the land shape by the use of GIS and to apply the information to the data mining algorithm C4.5, thereby suggesting a new method for addressing the problems of subjectivity and uncertainty. The findings showed approximately 81.6% agreement (21,570 parcels among the total 24,433 parcels) between the results of shape classification estimated with rules learned by C4.5 and the results of the existing surveys conducted by experts of Seoul City. Further, the kappa coefficient, which explains accidental or probabilistic agreement rate, was found to be 0.72, showing a considerable level of accuracy. The results of the experiment are significant in that the new method derived standardized rules that resolve the ambiguity of the existing survey. In particular, the method shows the effects of improvement in the consistency of characteristics information, which used to be classified variously according to surveyors. Given the inefficiency of the current land characteristics survey, which relies on manpower and field survey, it is deemed that the application of the computer simulation technique to the land characteristics survey will contribute to greater objectivity and efficiency in the land characteristics survey.