Comprehensive Detection of Fake Job Postings: Applying Advanced Data Balancing and Machine Learning Techniques to Combat Online Frauds

Fake job postings have become an increasingly common online crisis now-a-days, exploiting the money and identities of individuals seeking employment via online platforms. Genuine job postings are more in number compared to fake job postings; this imbalanced nature of the data makes it even harder to detect the fake job postings. This paper addresses the challenge of identifying fraudulent job postings by applying various data balancing techniques to counteract the class imbalance inherent in the dataset in combination with various state-of-art machine learning techniques. Methods like TF-IDF, categorical data are utilized for feature extraction and the impact of multiple balancing strategies, including SMOTE-NC (Synthetic Minority Over-sampling Technique for Nominal and Continuous), Cluster-Based Oversampling (Cluster-SMOTE) for data augmentation, and cost-sensitive learning is analyzed. After applying these techniques, we use several types of machine learning models on obtained balanced datasets. A comparative analysis of various machine learning models is performed alongside their corresponding data balancing techniques. Our results demonstrate the effectiveness of various balancing techniques in improving model performance, offering a more robust solution to combat online scams like fake job postings.