Investigating Fake Job Descriptions with TF-IDF and Word Embeddings

Fraudulent actions have increased as a result of the growth of online recruitment platforms, especially when it comes to the creation of false job descriptions intended to trick prospective employees. This study tackles the important problem of recognizing and categorizing fraudulent employment advertisements. By utilizing an extensive dataset of annotated fraudulent job descriptions, cutting-edge classification methods are utilized to construct a strong model that can discern between authentic and fraudulent job postings.TF-IDF, a feature extraction method is used to assign numerical weights to words based on their frequency. Several word embeddings, such as BERT, GloVe, and Word2Vec, were employed in this study to enhance the job descriptions’ feature representation for fraud detection. To overcome data imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) is also used in combination with TF-IDF. 5-fold cross validation is applied to the different classifiers to study the performance of the system. By merging numerous embeddings and resolving class imbalance, this ensemble model demonstrated a significant boost in fraud detection accuracy over extended testing, demonstrating the usefulness of doing so for robust categorization of misleading job descriptions.