Development of prediction model for linked data based on the decision tree - For track A, Task A1

Dongkyu Jeon, Wooju Kim

Research output: Contribution to journalConference article

1 Citation (Scopus)

Abstract

In this paper, we explain the detail analysis procedure of submission 1(Previous predicted results submission) of Task A1. We are trying to induce decision tree models to predict pc:numberOfTenders. Since the type of target attribute is non-negative integer value, we use the variance reduction as the at-tribute selection criteria. Input attributes are defined based on structure infor-mation of Public Contracts Ontology. We use the description logic constructors to properly represent a meaning of structure information of training data. Among all instances of the contract class, we make 10 different input data sets through random sampling method. The procedure of decision tree learning is performed by using SAS E-miner, and attribute selection criteria is variance re-duction. Final prediction results of test data are the average of selected decision tree models except few models which have extremely low R-Square value.

All Science Journal Classification (ASJC) codes

  • Computer Science(all)

Fingerprint Dive into the research topics of 'Development of prediction model for linked data based on the decision tree - For track A, Task A1'. Together they form a unique fingerprint.

  • Cite this