Identifying factors that influence student failure rate using Exhaustive CHAID (Chi-square automatic interaction detection)

Institutions hassles to accommodate a large of student that couldn´t passed in normal study period. Some of them pending the study period because couldn´t passed TPB in two semesters. If many students didn´t graduated on time, it would be a lot of difficulties involved by institutions. The impact of these problems such as human resources, supplying classroom and operational costs. This research tries to help stakeholders in the decision to reduce the impact by identifying factors that affecting the failure rate. The affecting factor could be determined by analyzing the courses taken and characteristic personal students. This research used Exhaustive CHAID as a method for building classification models to identify factors that influenced student failure rate. Those method is selected based on characteristic of dataset used in institutions records that consist of several type such as nominal, ordinal and floating. Besides the advantage of those method is could seek the most significant predictor variables by explore the structure dataset using the chi-square test statistic and p-value. This research has tested a different depth of tree and treshold (alpha_split) with different numbers of study program. Based on the experiment result, it shows that depth of tree can improve the accuracy but not significant with 85% average accuracy. It can be said that the method used can be considered.