The analysis of research fields of chinese natural science funds based on text classification

This paper applies text classification to mining research proposals for Chinese natural science funds. By using vector-space model to represent the proposals, the centroid-based classification has been carried out. And then the change of category members is analyzed through three variables defined to measure the stability, variability and overlapping of research fields, as the research field management science and engineering is taken as a case study. The resulting figures give a direct representation of characteristics of research fields, which are helpful to find out the interrelationship of research fields and to adjust the field code system to reflect the real status more accurately.