Klasterisasi Data Pajak Kendaraan Bermotor Menggunakan Model Hibrida Manhattan Frequency K-Means dan Particle Swarm Optimization

Febriansyah, Irfanur Ilham (2025) Klasterisasi Data Pajak Kendaraan Bermotor Menggunakan Model Hibrida Manhattan Frequency K-Means dan Particle Swarm Optimization. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 625202005-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Badan Pendapatan Daerah Provinsi Jawa Timur dituntut untuk terus melakukan optimalisasi Pendapatan Asli Daerah dari sektor Pajak Kendaraan Bermotor (PKB). Salah satu pendekatan yang dapat digunakan dalam mendukung analisis data PKB adalah klasterisasi, yaitu proses pengelompokan data berdasarkan karakteristik tertentu untuk menemukan pola tersembunyi dalam data. Penelitian ini menggunakan algoritma Manhattan Frequency K-Means (MFk-M) yang dinilai unggul dalam memproses data bertipe kategori secara efisien dan berskala besar. Untuk meningkatkan kualitas klaster yang dihasilkan, algoritma tersebut dikembangkan menjadi model hibrida dengan Particle Swarm Optimization (PSO) guna menentukan jumlah klaster yang optimal serta lokasi centroid terbaik secara otomatis dan lebih efisien. Model hibrida ini kemudian dibandingkan dengan tiga algoritma klasterisasi tunggal. Hasil eksperimen menunjukkan bahwa MFkM-PSO bersifat kompetitif dari sisi kualitas klaster, dengan nilai DBI dan Silhouette Score yang setara dibanding metode lainnya. Namun begitu dari sisi efisiensi waktu, model hibrida menunjukkan performa yang efisien terutama pada rentang jumlah klaster yang besar, menjadikannya pilihan yang unggul untuk skenario data berskala besar. Evaluasi kualitas klaster terhadap data PKB dilakukan dengan beberapa metrik evaluasi, yaitu Davies-Bouldin Index (DBI), Silhouette Score, Purity, Normalized Mutual Information (NMI), Entropy, serta waktu komputasi. Hasil penelitian ini diharapkan dapat memberikan model klasterisasi yang optimal dalam pengelompokan data PKB, serta menjadi dasar dalam pengambilan keputusan berbasis data secara lebih akurat dan sistematis.
=====================================================================================================================================
The Regional Revenue Agency of East Java Province is required to continuously optimize the region's own-source revenue, particularly from the Motor Vehicle Tax (PKB) sector. One of the approaches that can support PKB data analysis is clustering, which involves grouping data based on specific characteristics to uncover hidden patterns within the dataset. This study employs the Manhattan Frequency K-Means (MFk-M) algorithm, which is considered effective in processing categorical data efficiently and at scale. To improve the quality of the resulting clusters, the algorithm is enhanced into a hybrid model with Particle Swarm Optimization (PSO), which automatically and efficiently determines the optimal number of clusters and the best centroid positions. The hybrid model is then compared with three single clustering algorithms. Experimental results show that MFkM-PSO is competitive in terms of cluster quality, achieving comparable DBI and Silhouette Score values relative to other methods. In terms of computational efficiency, the hybrid model demonstrates superior performance, particularly when handling a large number of clusters, making it a favorable choice for large-scale data scenarios. The cluster quality evaluation on PKB data is conducted using several metrics, including Davies-Bouldin Index (DBI), Silhouette Score, Purity, Normalized Mutual Information (NMI), Entropy, and computational time. The results of this study are expected to provide an optimal clustering model for motor vehicle tax data grouping and serve as a foundation for more accurate and systematic data-driven decision-making.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Pajak Kendaraan Bermotor, Klasterisasi, Manhattan Frequency K-Means, Particle Swarm Optimization, Motor Vehicle Tax, Clustering, Manhattan Frequency K-Means, Particle Swarm Optimization
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	IRFANUR ILHAM FEBRIANSYAH
Date Deposited:	10 Nov 2025 03:42
Last Modified:	10 Nov 2025 03:42
URI:	http://repository.its.ac.id/id/eprint/128724

Actions (login required)

View Item