Implementation of data mining on customs false declaration detection

ADNANE FOUNOUN; MAHDI ALAOUI HANAF; AAWATIF HAYAR; ABDELKRIM HAQIQ

Author(s)

ADNANE FOUNOUN Hassan First University of Settat, Faculty of Sciences and Techniques, Computer, Networks, Mobility and Modeling laboratory: IR2M, 26000 -Settat, Morocco

MAHDI ALAOUI HANAF RITM Lab CED ENSEM/TICDev R&D Centre University Hassan Two ,Casablanca,Morocco.

AAWATIF HAYAR RITM Lab CED ENSEM/TICDev R&D Centre University Hassan Two ,Casablanca,Morocco.

ABDELKRIM HAQIQ Hassan First University of Settat, Faculty of Sciences and Techniques, Computer, Networks, Mobility and Modeling laboratory: IR2M, 26000 -Settat, Morocco

Abstract

Importation control in Indonesia is in need of an efficient method to oversee import declarations suspected of misleading information. This study offers a data mining method through Cross-Industry Standard Process for Data Mining (CRISP-DM) to provide a reliable method to predict false customs declarations. The method is tested using imports data from three major ports in Indonesia between 2019 and 2020. This study used decision trees algorithm as prior customs data mining study. The algorithm is compared with random forest and naïve bayes to detect false import declaration based on classification accuracy, precision, and recall scores. Testing results showed that random forest is the algorithm with the best classification accuracy, precision, and recall scores. Dataset preparation is very crucial since importation data has high cardinality and imbalance issues. This study used Synthetic Minority Oversampling Technique (SMOTE), normalization, and data transformation techniques to find a suitable dataset for a better prediction. The study finds that the model has a higher precision score than false declaration prediction based on whistleblower tips. This study is an addition to the limited amount of references pertaining to customs data mining.

Publication Details

Page(s) 1664-1674

DOI DOI not available

Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 6, Year: 2022

Request the full text from PASTIC

Keywords

Data Mining, SMOTE, CRISP-DM, Customs, False Declaration