Abstract
In Japan, cultivated land abandonment has been increasing rapidly in recent years and it causes degradation of biodiversity and decreases ecosystem services from agro ecosystem. Thus, it is required to determine the factors increasing cultivation abandonment and predict the future level of abandonment in regional scale in order to design strategy towards sustainable agriculture system under region-specific context. From this viewpoint, we tried to develop factor analysis and prediction model of abandonment cultivated land of Japan in reagional scale. We used census of agriculture and forestry distributed by Ministry of Agriculture, Forestry and Fisheries of Japan. And to identify the structure of the cultivation abandonment from the census data, we employed three machine learning algorithms, General Linear Model (GLM), Multivariate Adaptive Regression Spline and Random Forests (RF). The result shows that the best machine learning algorithm to predict the cultivation abandonment in Japan was RF algorism and it could detect major factors under reginal specific stituations.
