Utilization of Rule-based Predicting Models Faculty of Architecture, Urban Design and Creative Arts Students at Mahasarakham University by Data Mining

Abstract

Almost every technique in data mining can create effective rule-based prediction models. The purposes of this research are to study factors influencing selecting graduation bachelor and to create rule-based prediction models of graduation bachelor. The data were collected from 6 course education institutions in Faculty of Architecture, Urban Design and Creative Arts Students at Mahasarakham  University from year 2014-2016. The data consist of 929 records. In this paper, Decision Table, Decision Rule, Repeated Incremental Pruning to Produce Error Reduction (RIPPER) and OneR are used for building the rule-base prediction models. In order to measure the performance of the rule-base prediction models, 10-fold cross validation is applied to divide the data set into learning and testing sets. The performance of the prediction models created from each technique is measured with precision, recall and F-measure. The experimental results revealed that Factors influencing selecting graduation bachelor of the top five as follows: 1) district 2) Type of course 3) gender 4) grade point average of health and physical education and 5) Mother Career. For building prediction model, RIPPER is superior to Decision Table, Decision Rule, and OneR. It can create effective rule-based prediction models with 91%  of precision, 91.10%  of recall and 90.90%  of F-measure.