Detecting Fraud Job Recruitment Using Features Reflecting from Real-world Knowledge of Fraud

Abstract

A common method for text-analysis and text-based classification is to process for term-frequency or patterns of terms. However, these features alone may not be able to differentiate fake and authentic job advertisements. Thus, in this work, we proposed a method to detect fake job recruitments using a novel set of features designed to reflect the behavior of fraudsters who present fake information. The features were missing information, exaggeration, and credibility. The features were designed to represent in the form of a category and an automatically generatable score of readability. Data from EMSCAD dataset were transformed in accordance with the designed features and used to train a detection model for fake job detection. The experimental results showed that the model from the designed features performed better than those based on the term-frequency approach in every applied machine learning technique. The proposed method yielded 97.64% accuracy, 0.97 precision and 0.99 recall score for its best model when used for classifying fake job advertisements.