Selecting feature grouping and decision tree to improve results from the Learning Object Management Model (LOMM)

Recommendation systems, also known as intelligent decision support systems, have been used to support and strengthen the decision making in various areas including education. In order to establish efficient recommendation systems for educational purposes, several specific problems have to be addressed. One such problem is the weak relationship between input features, which causes performance of decision trees to deteriorate. This paper therefore proposes two preprocessing techniques to strengthen the relationships of input features for decision trees by using ontology and Apriori algorithm. Ontology-based feature grouping is used to combine related input features and to derive a set of new inputs. Apriori-based feature adding is used to find the groups of strong input features and add them as the new derived inputs. The proposed methods have been evaluated using data collected from schools in Nakhon Ratchasima province, Thailand. The experimental results suggested that the proposed methods have improved the accuracy of decision trees and the performance of recommendation systems in this test case. Furthermore, this paper also conducted experiments to select the appropriate input features and types of the decision tree specific to the dataset for further development.