ベイジアンネットワークを用いた緑視率に基づく緑環境評価の確率モデルに関する研究

抄録

 The purpose of this paper is to provide the method for the construction of the probabilistic model of the evaluation for the amount of green. The subjects of this study were the same as those used in the visible green study held in Kyoto City involving 61 landscape photos from 37 locations. Six elements of a green environment were defined, which include the ratio of visible green of a close-to-mid-range view, the ratio of visible green of a distant view, the other ratios of the sky, the pavement, the buildings, and the surface of water in a photo. Bayesian networks are expected to construct probabilistic models including an uncertainty of human behavior for prediction and decision-making. We applied Bayesian networks for providing a probabilistic model for the relationship between these elements and the evaluation for the amount of green.
 The graphical structure of the probabilistic model showed a direct relationship between the evaluation regarding the amount of green and the elements of close-to-mid-range and distant range of visible green. It also showed that the natural landscape elements, such as a river's surface, were related to satisfaction. The elements of close-to-mid-range views of visible green were found to be directly related to satisfaction regarding the amount of green, while the elements of the distant views were indirectly related.
 It was shown that the probabilistic model based on the probabilistic reasoning of Bayesian networks contained the relationship between the elements of the close-to-mid-range of visible green and the elements of the distant range, which was found in the visible green study in Kyoto City. It was also verified that satisfaction in regard to the amount of green tended to decrease with increasing ratios of buildings and pavements. Therefore, it was demonstrated that the probabilistic model did not deviate from the visible green study in Kyoto City and from predictable knowledge, verifying the model's validity.
 The probabilistic reasoning made by the model was used to predict the amount of visible green in close-to-mid-range views required for “satisfaction” and t.he evaluation of green. Such reasoning have been shown to match the sensitivity evaluation experiment. For the green environments are deduced to be “unsatisfactory, ” the green in close-to-mid-range views required for “satisfaction” were shown, demonstrating a way to improve the environments. Finally, a conclusion was made regarding the way to use the probabilistic model of green evaluation for application of the ratio of visible green.