Creating prediction models for the criminology licensure examination performance using JRip rule-based algorithm and J48 decision tree algorithm

The study explored the creation of prediction models for the Criminology licensure examination, identify significant predictors and compare the accuracy of the models generated. In the development of the models, the study used the JRip Rule-based and Decision Tree algorithms. In determining the significant attributes in predicting the performance of the criminology students, CfsSubsetEval was used as Attribute Evaluator and BestFirst was used as the Search Method. This was utilized to assess the value of a subset of attributes by considering each feature’s unique predictive ability as well as the degree of redundancy between them, and to search the space of attribute subsets using greedy hillclimbing augmented with a backtracking facility. The study revealed that the General Weighted Average (GWA) is a significant predictor. The Rule base generated six rules while the Decision tree generated five rules. It was revealed that Rule base model is accurate than the Decision tree with a class precision of 92.31. The models generated can be of help to the reviewer as it identifies students who needed special review assistance to eventually increase the licensure exam passing rate.