Real or Fake Job Posting Prediction using Random Forest, Long Short-Term Memory, and Multitational Naive Bayes

The rapid development of technologies has changed many aspects including job-seeking processes, enabling job seekers to explore job opportunities online. Unfortunately, this also led to the rising number of fraudulent job postings, deceiving individuals through misleading information. This research aims to build a model that accurately detects fake job postings, using machine learning/deep learning techniques, and to find key features that may lead to fraudulent job postings. The dataset is taken from Kaggle where exploratory data analysis and preprocessing is done, such as cutting data entries, cleaning, stemming/lemmatization, tokenization, vectorization, and resampling to handle class imbalance. Three algorithms, Random Forest, Long Short-Term Memory (LSTM), and Multinomial Naive Bayes (MNB), were implemented with various resampling methods. The models were evaluated based on its accuracy and F1-Score. Results revealed that Random Forest paired with Adaptive Synthetic (ADASYN) resampling method outperformed the other models, achieving 96% accuracy and a high F1-score of 0.87. These findings highlighted common characteristics of fraudulent job posts, such as lacking or vague company profile, description, and requirements. Future work may involve exploring more features and expanding the dataset to enhance model performance.