Predict Rhododendron L. Potential Distribution Center In China By SVM

Rhododendron L. is the biggest genus of spermatophyte in China. Its current distribution and differentiation center is the southwest of China, including Hengduan Montain and eastern Himalayas. In west and southwest of China, Yunnan Province, Guizhou Province and Sichuan province, there are more than 450 species of Rhododendron L.and about 300 of them are endemic species of China. Therefore, it is important to study further distribution of Rhododendron L. for biodiversity conservation of Hengduan Montain.The most common method to build a predictive model of species potential distribution is to use environmental factors, because they tightly affect the species distribution. Unfortunately, most predictive models suffer from the"high dimension small sample size"problem—cannot give satisfactory result when there is only liminted specimen data, and cannot handle large number of environment factors. Support Vector Machine (SVM), which is based on Structural Risk Minimization principle, has been proved to be especially suitable for such kind of data by both theory and abandon applications in machine learning domain. Here, we implement a new predictive system of species potential distribution based on SVM method. In order to investigate the effectiveness of the method, we perform a country-scale case study using 30 species of Rhododendron L. in China with their specimen data and 11 layers of 1km digital environmental grid-data. Through expert evaluation and Receiver Operator Characteristic (ROC) curve, I compare SVM with the commonly used Genetic Algorithm for Rule-Set Prediction (GARP). Our experiment shows that all SVM's predictions are consistently better than GARP's in terms of expert's evaluation. For the statistical analysis of ROC curve, almost all the Area Under the...