A Novel Online Job Scam Detection of Imbalanced Data Using ML and NLP Models

Abstract

This research initiative addresses the pervasive threat of online job recruitment scams by leveraging a potent machine learning model fortified with natural language processing (NLP). While the internet expands job search horizons, it concurrently exposes job seekers to fraudulent practices, enticing them with false opportunities and extracting sensitive information or money. This work will adhere to the CRISP-DM methodology. Through the implementation of varied machine learning algorithms such as random forest, support vector classifier, Gaussian Naive Bayes, LightGBM, and XGBoost in conjunction with natural language processing models like Uni-Gram, Bi-Gram, Tri-Gram, and TF-IDF using the balanced data set, it was discovered that the Gaussian Naïve Bayes model performed the best for both trigram and TF-IDF using random under sampling and oversampling.