A prototype of fake job posting prediction system using logistic regression

Online job platforms make job searching efficient by offering broader opportunities. However, scammers use these platforms to deceive job seekers by creating fake job postings. Fraudsters use complex tactics that make it hard to differentiate between legitimate and fraudulent posts. It can be harmful for job seekers as it risks their privacy and also financial. The objectives of this system is to develop a prototype of fake job posting prediction using logistic regression and to evaluate the performance of the logistic regression algorithm. This Fake Job Posting Prediction System uses a Logistic Regression trained on a Employment Scam Aegean Dataset(EMSCAD) of job postings. Key preprocessing steps include handling missing values, combining important text fields, encoding categorical variables, and applying TF-IDF vectorization. SMOTE was used to balance the dataset, and GridSearchCV was applied for hyperparameter tuning. The model achieved an accuracy of 97.83%, recall of 79.49%, precision of 76.54% and an F1-score of 77.99% using a 90/10 train-test split with SMOTE and tuning applied. It successfully predicts fake job postings and provides keyword-based explanations to support user understanding. The proposed system successfully predicts fake job postings using a transparent and user-friendly interface.