Deteksi Penipuan Rekrutmen Kerja Menggunakan Embedding Bidirectional Encoder Representations From Transformers (BERT)

Salma, Alya Putri (2025) Deteksi Penipuan Rekrutmen Kerja Menggunakan Embedding Bidirectional Encoder Representations From Transformers (BERT). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025211174-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Penipuan yang melibatkan rekrutmen kerja menjadi masalah signifikan yang dapat merugikan para pencari kerja. Di tengah berkembangnya perekrutan online, risiko penipuan juga semakin meningkat, di mana pelaku kejahatan siber membuat iklan rekrutmen palsu dengan tawaran pekerjaan menggiurkan. Penelitian ini bertujuan untuk mengidentifikasi rekrutmen pekerjaan palsu menggunakan metode embedding BERT, yang kemudian diklasifikasikan dengan beberapa metode, yaitu CNN, LSTM, dan CNN-LSTM. Penelitian ini juga mengatasi masalah ketidakseimbangan data dengan menggunakan teknik Random Oversampling dan Random Undersampling. Hasil pengujian menunjukkan bahwa model BERT + CNN-LSTM dengan kombinasi konfigurasi terbaik dari model CNN dan LSTM, yaitu satu layer CNN dengan jumlah filter 256 dan satu layer LSTM dengan hidden size 64, memberikan performa terbaik dengan teknik Random Oversampling pada rasio pembagian data 80:10:10. Model ini mencatatkan accuracy 99,27%, precision 96,30%, recall 95,80%, dan F1-score 96,05%. Hasil ini menunjukkan bahwa konfigurasi tersebut mampu mendeteksi rekrutmen kerja palsu dengan tingkat akurasi yang optimal.
======================================================================================================================================
Fraud involving job recruitment has become a significant issue that can harm job seekers. With the rise of online recruitment, the risk of job recruitment fraud has also increased, as cybercriminals take advantage of job seekers' vulnerabilities by posting fake job advertisements promising attractive positions and high compensation. These frauds target identity theft, financial scams, and other malicious activities. This research aims to identify whether job recruitment is fraudulent using the BERT embedding method and classifies it using several models, namely CNN, LSTM, and CNN-LSTM. The study also addresses the issue of imbalanced data by applying Random Undersampling and Random Oversampling techniques. The best performance was achieved using the BERT + CNN-LSTM model with the Random Oversampling technique on a data split ratio of 80:10:10. This model combined the optimal configurations from both CNN and LSTM, with one CNN layer and 256 filters, and one LSTM layer with a hidden size of 64. The model achieved an accuracy of 99.27%, precision of 96.30%, recall of 95.80%, and F1-score of 96.05%. These results demonstrate that this configuration is optimal for detecting fraudulent job recruitments with high accuracy.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	BERT, CNN, Confusion Matrix, Penipuan Kerja, BERT, CNN, Confusion Matrix, Job Fraud
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering
Depositing User:	Alya Putri Salma
Date Deposited:	28 Jul 2025 08:10
Last Modified:	28 Jul 2025 08:10
URI:	http://repository.its.ac.id/id/eprint/122209

Actions (login required)

View Item