A Hybrid Evaluation Index Approach in Optimizing Single Tuition Fee Cluster Validity

The grouping of the socio-economic level of new students at the time of registration at public universities is a problem faced by all state universities. Identifying the right group will have an impact on the students and the university. The quality of the results of a valid grouping will give a sense of fairness to the parents of students in paying tuition fees. On the other hand, the university also expects that the results of a valid grouping will contribute to optimal revenue. This study aims to evaluate the cluster structure of a single tuition fee at the State University of Surabaya. The existing cluster structure is compared with the results of grouping using nine clustering methods, namely K-Mean, Hierarchical, BIRCH, DBSCAN, Mini Batch K-Mean, Mean Shift, OPTICS, Spectral Clustering, and Mixture Gaussian. The proposed evaluation method is a combination of three evaluation concepts, namely internal validity (Silhouette-Index), external validity (Rand Index), and the percentage conformity value to the expected income factor (Revenue-Index). These three indicators are then calculated as the average value for each clustering method as Hybrid-Index. The highest Hybrid-Index is shown by the Mini Batch K-Mean algorithm, with an average value of 0.6420, so the Mini Batch K-Mean algorithm can be recommended as a method for grouping single tuition fees.