TY - GEN
T1 - Data Clustering for Tax Incentives Determination
T2 - 1st International Conference on Artificial Intelligence Technology, ICoAIT 2025
AU - Febriansyah, Irfanur Ilham
AU - Amaliah, Bilqis
AU - Saikhu, Ahmad
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Motor Vehicle Tax (PKB) plays a crucial role in financing regional development in Indonesia. However, the application of data science to support the determination of tax rates and related incentives remains limited. This study implements a hybrid approach that combines Manhattan Frequency K-Means (MFKM) and Particle Swarm Optimization (PSO) to automatically determine the optimal number of clusters for large-scale, categorical PKB datasets. The novelty lies in applying this combination - one of the earliest implementations of MFKM-PSO in large-scale tax data analysis in Indonesia - to a complex fiscal challenge, where existing methods struggle with both accuracy and efficiency. The resulting clusters are then interpreted to inform the design of more targeted and data-driven PKB incentive policies. Using 208,157 PKB records, the proposed method successfully identified three optimal clusters, supported by a Davies-Bouldin Index (DBI) of 0.9072 and a Silhouette Score of 0.5312, within 38.53 seconds of computation time. Cluster interpretation and simulation of a rule-based incentive policy revealed potential for a 2.87% reduction in PKB loss, indicating practical relevance for fiscal decision-making.
AB - Motor Vehicle Tax (PKB) plays a crucial role in financing regional development in Indonesia. However, the application of data science to support the determination of tax rates and related incentives remains limited. This study implements a hybrid approach that combines Manhattan Frequency K-Means (MFKM) and Particle Swarm Optimization (PSO) to automatically determine the optimal number of clusters for large-scale, categorical PKB datasets. The novelty lies in applying this combination - one of the earliest implementations of MFKM-PSO in large-scale tax data analysis in Indonesia - to a complex fiscal challenge, where existing methods struggle with both accuracy and efficiency. The resulting clusters are then interpreted to inform the design of more targeted and data-driven PKB incentive policies. Using 208,157 PKB records, the proposed method successfully identified three optimal clusters, supported by a Davies-Bouldin Index (DBI) of 0.9072 and a Silhouette Score of 0.5312, within 38.53 seconds of computation time. Cluster interpretation and simulation of a rule-based incentive policy revealed potential for a 2.87% reduction in PKB loss, indicating practical relevance for fiscal decision-making.
KW - Hybrid Clustering
KW - Large-Scale Dataset
KW - Manhattan Frequency K-Means
KW - Motor Vehicle Tax
KW - Particle Swarm Optimization
KW - Tax Incentives Simulation
UR - https://www.scopus.com/pages/publications/105031900542
U2 - 10.1109/ICoAIT67446.2025.11308916
DO - 10.1109/ICoAIT67446.2025.11308916
M3 - Conference contribution
AN - SCOPUS:105031900542
T3 - ICoAIT 2025 - 1st International Conference on Artificial Intelligence Technology - Artificial Intelligence: Driving Prosperity and Sustainability in the Modern World
SP - 83
EP - 88
BT - ICoAIT 2025 - 1st International Conference on Artificial Intelligence Technology - Artificial Intelligence
A2 - Pranolo, Andri
A2 - Ismi, Dewi Pramudi
A2 - Pawestri, Sheraton
A2 - Khoirunnisa, Ninda
A2 - Ismail, Amelia Ritahani
A2 - Snani, Aissa
A2 - Wicaksono, Hendro
A2 - Abdalla, Modawy Adam Ali
A2 - Voliansky, Roman
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 10 September 2025 through 11 September 2025
ER -