<![CDATA[Biclustering is a two-dimensional clustering approach that simultaneously identifies relationships between objects and variables, making it more suitable than conventional clustering methods for analyzing data with local patterns. Although numerous biclustering algorithms have been developed, comparative studies evaluating their performance on official statistical data remain limited. This study aimed to evaluate the performance of the Cheng and Church (CC) and Qualitative Biclustering (QUBIC) algorithms using Indonesia’s aquaculture production data for 2023. The dataset comprised the production of 14 aquaculture commodities across 34 provinces. Optimal parameters for both algorithms were determined through a manual tuning procedure. Their performance was subsequently evaluated using the Mean Squared Residue (MSR), Average Squared Residue (ASR), and the Liu–Wang index. The results showed that the optimal parameter for the CC algorithm was δ = 0.009, while the optimal QUBIC parameters were r = 1, q = 0.06, and c = 0.75. CC produced 7 biclusters with an ASR of 0.0052, whereas QUBIC produced 8 biclusters with an ASR of 0.4360. The lower MSR and ASR values of CC compared with QUBIC indicate that CC was more effective in forming homogeneous biclusters in aquaculture production data. The number of province memberships showed that both algorithms were able to capture patterns across most provinces, although CC had slightly broader coverage. The Liu and Wang index value of 0.1951 indicated a relatively low similarity between the bicluster membership structures produced by CC and QUBIC. Thus, although both algorithms were applied to the same aquaculture production data, CC and QUBIC tended to produce different clustering structures.]]>
Copyrights © 2026
