Evaluation of Bicluster Analysis Results in Capture Fisheries Using the BCBimax Algorithm

Abstract

Biclustering is a simultaneous clustering technique by finding sub-matrixes that have the same similarity between rows and columns. One of the biclustering algorithms that is relatively fast and can be used as a reference for the comparison of several algorithms is the BCBimax algorithm. The BCBimax algorithm works by finding a sub-matrix containing element 1 of the formed binary data matrix. The selection of thresholds in the binarization process and the minimum combination of rows and columns are essential in finding the optimal bicluster. Capture fisheries have an important role in supporting sustainable growth in Indonesia, so information on the potential of fish species that have similarities in several provinces is needed in optimally mapping the potential. The BCBimax algorithm found 11 optimal biclusters in grouping capture fisheries data. The median of each variable is used as a threshold in the binarization process, and the minimum combination of row 2 and maximum column 2 is chosen to find the optimal bicluster result. The optimal average value of Mean Square Residual bicluster obtained is 0.405403 with the similarity of bicluster results (Liu and Wang index) which is different for each bicluster combination produced. All the bicluster results grouped the provinces and types of fish that had the same potential simultaneously.

Pattern Recognition of Food Security in Indonesia Using Biclustering Plaid Model

“…Secondly, it is based on the smallest AvMSR value. Furthermore, another important consideration is how the information obtained from the biclustering results (Wulandari et al, 2023), in other words the model can produce more than one bicluster (𝐵𝐶 > 1). Based on Table 3, it can be shown that models 1 and 3 have 𝑟 value less than 1 (𝑟 < 1).…”