Fraudulent Job Posting Classification Using AI

Abstract

Online recruitment platforms are becoming increasingly indispensable in modern employment markets, offering millions of listings across various industries and geographies. However, these platforms are increasingly exploited by malicious actors who publish fraudulent job postings for scams, including identity theft, monetary fraud, and data harvesting. These fraudulent postings not only affect job seekers but also damage the reputation and credibility of online recruitment portals. Manual verification of postings is infeasible at scale, creating the urgent need for automated detection systems. This study presents a machine learning pipeline for fake job posting detection that integrates preprocessing, feature engineering, class balancing, and ensemble learning. A publicly available dataset of 18,000 postings was preprocessed to address missing values, normalized to reduce inconsistencies, and augmented with engineered features such as word counts and binary attributes. Textual data were vectorized using TF-IDF, while categorical attributes were encoded using one-hot encoding. Class imbalance was resolved using the synthetic minority oversampling technique (SMOTE). The final model, a random forest classifier, achieved 95% accuracy, 94% balanced accuracy, and a ROC-AUC of 0.98. A confusion matrix and ROC curve validate the system’s ability to correctly identify fraudulent postings with minimal false positives. The pipeline was deployed as a Flask-based web application for real-time prediction.