Support vector machines (SVMs), which are based on statistical learning theory, have recently got considerable potential
in data mining by their good ability of generalization. Land grade evaluation, which provides information for land
planning and decision-making, is a process of evaluating land quality for a particular use. It can be referred to as
multiclass classification problem. As a result of SVMs' binary nature, they can not be directly applied to land grading
process. By integrating SVMs with a binary tree, this paper applied a binary tree based SVMs (BTSVMs) approach into
land grade evaluation. Arable land in Heping County, Guangdong, was chosen as study area and BTSVMs model was
then applied to the data. In addition to BTSVMs, the same data were classified using decision tree (DT) and artificial
neural network (ANN). Compared with DT and ANN, results showed that BTSVMs had better classification accuracy.
While decreasing the size of training data, the accuracy of each approach dropped down positively with BTSVMs
relatively more accurate than others. In general, BTSVMs is potentially feasible in the application of land grade
evaluation with its good performance.
|