Fake Job Vacancy Detection Using Ensemble Voting Classifier

The increase in fake job vacancy information has become a concern for people looking for jobs in the current digitalization era. Survey shows that 79% of job seekers search for job vacancies on the internet. The high demand is being exploited by malicious people, resulting in identity leaks and material losses for job seekers. This research is conducted to create a system that can detect fake job vacancies on the internet using techniques and algorithms that are increasingly developing. Author used several machine learning approaches such as Random Forest (RF), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Extra Trees (XT), and also voting classifier to perform classification on fake job vacancies. In addition, author also use Natural Language Processing (NLP) to perform data cleaning, Term Frequency-Inverse Document Fre- quency (TF-IDF) for feature extraction, and Synthetic Minority Oversampling Technique (SMOTE) for oversampling to overcome data imbalance. After implementing these approaches, the single classifier achieves F1 values between 60-75%. Meanwhile, the designed voting classifier model using XT, SVM, and XGB as base estimators can produce an accuracy of 98.21%, F1 of 77.7%, and MCC of 76,32%. In conclusion, the voting classifier model with SVM, XGB, and XT estimators can effectively improve the quality and accuracy of the predictions produced.