Clustering Mixed-Type Data on The Scientific Publication Productivity of ITS Lecturers Using Cluster Ensemble Based Mixed Data Clustering (CEBMDC)

The disparity in scientific publication productivity among lecturers at Institut Teknologi Sepuluh Nopember (ITS) poses a critical challenge to achieving its goal of becoming a World Class University (WCU). This issue is compounded by the complexity of the available data, which consists of mixed types data. To address this challenge, this study has two main objectives: to identify the most effective method for clustering lecturers based on their scientific publication productivity, and to compare the performance of the CEBMDC DBSCAN-QROCK and CEBMDC HDBSCAN-QROCK algorithms when applied to mixed-type data containing outliers. The clustering performance is evaluated using the SW/SB ratio for ensemble results. The CEBMDC DBSCAN-QROCK configuration yields the best clustering quality, achieving an SW/SB ratio of 0.711 and producing six distinct clusters with one outlier. The analysis reveals a stark gap in productivity, where only a small segment of lecturers demonstrates high performance, while the majority are still developing. This insight supports the implementation of targeted development programs, such as cluster-based mentoring and cross-faculty collaboration. This research is the first known application of an ensemble clustering method to lecturer productivity data in ITS. Future work may focus on extending the ensemble framework with adaptive thresholds or applying it to other performance-based clustering problems in academia.