Abstract:

The admission processes in higher education can be conducted in many ways such as using standardized testing or interviewing. After candidates pass these particular processes, they will become freshmen in a university; however, some of them suffer from several problems that cause them to drop-out of the university later on. During the admission process, demographic data from the applicants were collected. Hence, if we can utilize the data by analyzing it to find the factors that affected the students’ study in their universities; the university, relevant persons, and the students could be managed and their study plan can be appropriately improved for the future. According to previous research concerning the demographic data, we found that much statistical research focuses on affected factors of education. The results found that GPA and parent income were affected by education. Moreover, the study about demographic and psychological factors influencing academic success in a college-level human anatomy course confirmed that demographic data affected education in differing ways. Other research applied the Associative Classification technique to classification and prediction by using the association rule with the training and testing approach for group prediction. As this particular technique was effective in many issues, it was one of the techniques that the researcher used to evaluate the freshmen quality from the demographic data. This research purpose was to create a classifier tool to evaluate freshmen’s ability.The research also used demographic data from students of the Information Technology program at Chandrakasem Rajabhat University and their current grade levels as equalized levels in the testing. The equalized levels consist of 3 classes: good, fair and poor. The procedure of PFSCARs is comprised of: 1) data preparation –to prepare student demographic data from a sample of 1,003 students, transform some continuous attributes to an appropriated format and separate it into a training set (668 records) to generate CARs and a testing set (335 records) for accuracy estimates 2) generate PFSCARs –This process is to generate the longest CARs, called Particular Full-Scaled CARs, by all frequent rule items which pass the MinSupp and MinConf threshold. In this study we specify MinSupp as 0.01% and MinConf as 45%. 3) classification –choose PFSCARs by examining the precedence already used as a student classifier tool and the class label from the most precedence rules will be selected as results. The PFSCARs in this research found 52 good class label rules, 110 fair class label rules and 104 poor class label rules that were composed overall of 266 PFSCARs. However, within MinSupp and MinConf there were 157 equalized PFSCARs consisting of 3 class label rules 43, 61 and 40 as good, fair and poor respectively. From these, it can be clearly seen that the forecasting model to evaluate freshmen quality with PFSCARs performed at a good level of performance with an accuracy rate of 79% of the students equalized model. The restriction of this study is the lack of a data set in good and poor class labels that affected its accuracy performances. Finally, the research discovered that the forecasting model to evaluate freshmen quality can be a guideline for academic advisors or other relevant persons to help new students, to manage an appropriated study plan and could help them to improve course or curriculum in the future as well.

Keywords:

freshmen s ability, class association rules, forecasting.