Pemanfaatan Artificial Intelligence untuk Optimalisasi PNBP: Studi Kasus Bea Lelang pada Lelang Indonesia

Sujak, Galuh Mafela Mutiara
Rofiq, Hanif Noer

Indonesian Treasury Review: Jurnal Perbendaharaan, Keuangan Negara dan Kebijakan Publik
Vol 9 No 2 (2024): Indonesian Treasury Review: Jurnal Perbendaharaan, Keuangan Negara dan Kebijakan

Misclassification of auction objects can result in an inaccuracy of the Auction Fee that is imposed, resulting in under/overpayment of government revenue, a decline in public reputation, and differences in auction fee data in SIMPONI and Portal Lelang Indonesia. These errors can be anticipaed by adding verification step by the Auctioneer. Meanwhile, the increase in the frequency of auctions is disproportionate to the number of Auctioneer, thus a mechanism that can assist the Auctioneer to do verification without adding additional work is needed. The authors propose the use of a Convolutional Neural Network to carry out the automatic classification of auction objects in the form of Buildings, Demolition, Cars, and Motorcycles. The dataset was obtained from the Portal Lelang Indonesia. The results of training and validation accuracy were 96.13% and 96.50%. The model is then applied to a dashboard for manual testing, and 100% accuracy results are obtained from all the images tested.