The environment of higher education has been changing constantly. As national and social demands for quality management in higher education increase, new ways such as big data technology to maximize the value of quality management in higher education also emerge. Under this circumstance, it is necessary to promote a model that can provide meaningful useful information using big data on higher education. The purpose of this study is to develop predictive model of quality management in higher education using data mining and AHP based on Higher Education Information Disclosure data.
Three research questions were investigated: 1) What is AHP model derived from the analysis of previous studies and literature review? 2) What is relative weight of each variable calculated through AHP? 3) What is a predictive model of quality management in Higher Education derived through data mining?
To answer these questions, this study conducted literature review, AHP survey, and data mining. First, the components of higher education were explored through literature review. Secondly, AHP survey was conducted from experts in higher education, and relative weights and priorities of the variables were computed. Finally, based on the results of AHP analysis, input variables up to top 15 were selected as predictors, and a predictive model of quality management higher education for 7 target variables was developed using decision tree and random forest, respectively.
The results showed as follows: First, the input AHP model constructed 5 factors, 31 indicators and the output AHP model constructed 2 factors, 11 indicators. Second, the AHP analysis was used to assess the weight and priority of each of factors and indexes. 'Curriculum'(0.28) was the highest factor, followed by the 'University finance' and 'Educational expenditure'(0.25), 'Condition of education'(0.22), 'Student'(0.20), and 'Internationalization'(0.05) in the input AHP model. When calculating global priority of input variables, the three indicators with the highest priority were the 'educational restitution rate', 'cost of education per student', and 'retention rate of full-time faculty'. It means that experts in higher education considered financial and human resources that institutions actually invested in supporting students' educational environment, and practical curriculum as significant factors.
For output factors, 'education outcome'(0.72) appeared to be the highest weight factor and 'research outcome'(0.28) followed. For global priority analysis on 11 output indicators, the 'employment rate' and 'recruitment rate' were found to be the most important.
Third, the result of data mining(decision tree, random forest) using the top 15 input variables based on data from 167 colleges showed all target variables, excluding 'employment rates', were influenced by the 'amount of governments financial program', 'educational restitution rate', and 'cost of education per student'. 'Employment rate' was related to 'percentage of industry full-time faculty', 'filed training curriculum', and 'percentage of full-time faculty'. In addition, the result of deriving regional models indicated that the 'amount of governments financial program' and 'educational restitution rate' appeared to have major impact in most regions. However, the variables that have major influences on the target variable were somewhat different depending on the characteristics of the colleges from different regions.
Based on these results, this study suggest that the government needs to draw up financial support plan diversification to improve the educational quality in higher education. Also, government should consider the validity of college evaluation criteria. For example, the employment rate needs to be reconsidered for college assessment. Finally, colleges from different regions showed different conditions and characteristic in this study, thus, it is necessary to review the entire evaluation process including regional classification and methods of conducting evaluation by types with college size and attributes.