Penggerombolan Perusahaan Budidaya Ikan di Indonesia dengan Metode K-Prototypes, Two Step Cluster (TSC), dan Gower
Abstract
Subsektor perikanan memegang peranan yang penting terhadap perekonomian Indonesia. Perikanan dibagi menjadi dua, yaitu perikanan tangkap dan perikanan budidaya. Namun, perikanan budidaya lebih mendominasi terkait produksi perikanan nasional dibanding perikanan tangkap. Sumber produksi utama perikanan budidaya yaitu melalui perusahaan budidaya ikan. Pada tahun 2018 terdapat 258 perusahaan budidaya ikan berbadan hukum yang aktif di Indonesia dan diperoleh melalui Laporan Tahunan Perusahaan Budidaya Ikan yang dilakukan oleh Badan Pusat Statistik (BPS). Setiap perusahaan budidaya ikan memiliki karakteristik yang berbeda-beda, terutama mengenai keadaan perekonomian dan perkembangan usahanya, sehingga perlu dilakukan penggerombolan untuk lebih mudah memahami kondisi perusahaan-perusahaan tersebut melalui karakteristik gerombol yang dihasilkan. Salah satu metode statistik yang dapat dilakukan yaitu analisis gerombol (cluster analysis). Analisis gerombol klasik biasanya hanya dapat digunakan untuk pengolahan satu tipe data saja, yaitu tipe data numerik saja atau tipe data kategorik saja. Penelitian ini melibatkan 21 peubah, yaitu 8 peubah kategorik dan 13 peubah numerik. Oleh karena itu diperlukan metode penggerombolan yang tepat untuk melakukan penggerombolan perusahaan budidaya ikan di Indonesia. Beberapa metode yang dapat digunakan, yaitu K-Prototypes, Two Step Cluster, Metode Gower, Cluster Ensemble, dan Latent Class Cluster. Namun, penelitian ini hanya membatasi penggunaan dari tiga metode penggerombolan, yaitu K-Prototypes, Two Step Cluster (TSC), dan Gower. Hasil menunjukkan bahwa metode K-Prototypes merupakan metode penggerombolan perusahaan budidaya ikan terbaik dibanding metode TSC dan Gower. Hal ini dikarenakan metode ini menghasilkan nilai rasio antara keragaman di dalam gerombol dan keragaman antar gerombol yang paling kecil. Jumlah gerombol perusahaan budidaya ikan yang terbentuk yaitu sebanyak 6 gerombol. Jumlah anggota gerombol 1 sebanyak 8 anggota, gerombol 2 sebanyak 65 anggota, gerombol 3 sebanyak 19 anggota, gerombol 4 sebanyak 75 anggota, gerombol 5 sebanyak 14 anggota, dan gerombol 6 sebanyak 77 anggota. The fisheries subsector plays an important role in the Indonesian economy. Fisheries are divided into two, namely capture fisheries and aquaculture fisheries. However, the aquaculture fisheries dominates more than the capture fisheries in terms of national fisheries production. The main source of aquaculture production is through the aquaculture fisheries companies. In 2018 there were 258 aquaculture fisheries companies with legal entities that were active in Indonesia and they were obtained through the Annual Report of Aquaculture Fisheries Companies that conducted by the Statistics Indonesia (BPS). Each aquaculture fisheries company has different characteristics, especially regarding to the state of the economy and the development of its business, so clustering is necessary to more easily understand the conditions of these companies through the characteristics of the resulting clusters. One of the statistical methods that can be done is cluster analysis. Usually the classical cluster analysis only can be used for processing one data type, namely numerical data types or categorical data types. This study involved 21 variables, 8 categorical variables and 13 numerical variables. Therefore, appropriate clustering methods are needed to cluster the aquaculture fisheries companies in Indonesia. Several methods can be used, they were K-Prototypes, Two Step Cluster (TSC), Gower Method, Cluster Ensemble, and Latent Class Cluster. However, this study only limits the use of three clustering methods, namely K-Prototypes, Two Step Cluster (TSC), and Gower. The results show that the K-Prototypes method is the best clustering method for aquaculture fisheries companies compared to the TSC and Gower methods. This is because this method produces the smallest value of the ratio between the standard deviation within the cluster and the standard deviation between clusters. The number of clusters of aquaculture fisheries companies that formed is 6 clusters. Cluster 1 has 8 members, cluster 2 has 65 members, cluster 3 has 19 members, cluster 4 has 75 members, cluster 5 has 14 members, and cluster 6 has 77 members.