Model prediction for accreditation of public junior high school in Bogor using spatial decision tree

Abstract

Indonesia has a large geographic area with large variance of quality service of education. This paper will analyze whether it has a correlation or not, between the quality level of school and its characteristic of geographic area. This paper describes performance of two kind of decision tree method in predicting level of accreditation for public junior high school in Bogor, West Java. With three scenarios (training set as testing set, leave one out, and 3-fold cross validation) using conventional decision tree (ID3) the accuracy level reach 97.14%, 40%, and 51.51% respectively. On the other hand, using spatial decision tree (SDT) the accuracy reach level 97.14%, 54.29%, and 54.28%. Based on the accuracy level and the structure of tree that was constructed, SDT produce better result. These facts will imply that quality level of public junior high schools on Bogor have correlation with characteristics of geographic area. However, SDT need much more computation time to construct decision tree rather than ID3, so for big number of data and big number of attributes, SDT will not appropriate.