Developing debt risk model using data mining for IRBM / Hamisah Mohamed Azidin

This paper describes the development of a debt risk predictive model for individual taxpayer in Inland Revenue Board Malaysia (IRBM). Using data mining to predict taxpayer's compliance and non-compliance has gained attention in recent times. However, very little research has been done to predict taxpayers who have debt with tax organizations around the world. The objective of this study is to choose a suitable data mining methodology or framework and what is suitable data mining technique to build a predictive model debt risk for taxpayers who have debts with IRBM. This study also to get the behavior or pattern of data and identify important variables in predicting taxpayer with debt risk. Data individual taxpayers who have debt value until 31.12.2013 obtained from the database data warehouse and data mart IRBM. These data were analyzed using IBM SPSS version 16.0 applications. From the study of data mining framework and technique in theory, we choose CRISP-DM data mining life cycle framework include business understanding, data preparation, build models, evaluation and test the model as a framework and CHAID decision tree as technique. The results showed that the method CHAID Decision Tree in the model built to the study of type of test the value of 80% accurate. The result revealed that he is accurate model to use to predict future taxpayers which have outstanding debt with IRBM.