JOBSHIELD ANALYTICS: COMPARING MACHINE LEARNING APPROACHES IN FRAUD DETECTION

Abstract

In recent times, the proliferation of modern technology and widespread social
communication has led to a surge in job advertisements, making the detection of fake
job postings a critical concern. Predicting the authenticity of job posts poses
substantial challenges in the realm of classification tasks. This study proposes
leveraging various data mining techniques and classification algorithms, including KNearest
Neighbors (KNN), Decision Tree, Support Vector Machine (SVM), Naïve
Bayes Classifier, Random Forest Classifier, Multilayer Perceptron, and Deep Neural
Network (DNN), to discern whether a job post is genuine or fraudulent. The
experimentation is conducted on the Employment Scam Aegean Dataset (EMSCAD),
comprising 18,000 samples. Notably, the Deep Neural Network emerges as a
formidable classifier, exhibiting exceptional performance in this classification task.
The employed DNN architecture comprises three dense layers, achieving an
impressive classification accuracy of approximately 98% in predicting fraudulent job
posts. The contemporary surge in job postings, fueled by advancements in technology
and widespread social communication, has brought forth a pressing concern—
detecting fraudulent job posts. This project presents a comprehensive comparative
study on the detection of fake job posts, employing various machine learning algorithms. The classification algorithms under scrutiny include K-Nearest Neighbors
(KNN), Decision Tree, Support Vector Machine (SVM), Naïve Bayes Classifier,
Random Forest Classifier, Multilayer Perceptron, and Deep Neural Network (DNN).
The empirical evaluation is conducted on the Employment Scam Aegean Dataset
(EMSCAD), comprising 18,000 samples. Our findings reveal that the Deep Neural
Network (DNN) emerges as a standout classifier, showcasing remarkable
performance with an accuracy of approximately 98% in predicting fraudulent job
posts. The comparative analysis sheds light on the strengths and weaknesses of each
algorithm, providing valuable insights into their efficacy for fake job post detection.
This study contributes to the ongoing discourse on leveraging machine learning
techniques to address the escalating challenges posed by deceptive job postings in the
contemporary employment landscape.