Real or Fake Job Posting Detection

Abstract

This research presents a machine learning approach to distinguish between legitimate and fraudulent job postings in the recruiting sector. The dataset used, labelled as 'authentic list,' comprises approximately 17,880 entries from Kaggle and includes various attributes such as job title, location, salary range, company profile, job description, industry, and indicators of fraudulent activity in job advertisements. The proposed methodology begins with Exploratory Data Analysis (EDA) to gain insights into the multi-class classification of different features and to identify correlations within the dataset. Data pre-processing techniques, including Natural Language Processing (NLP), are employed to prepare the datasets for training and testing. Several machine learning algorithms such as K-Nearest Neighbours (KNN), Support Vector Machine (SVM), Random Forest, Logistic Regression, Naive Bayes, and AdaBoost are used to classify job listings as legitimate or fraudulent. The performance of each classifier is evaluated using qualitative metrics such as accuracy, precision, recall, F1-score, selectivity, and specificity. The results show the effectiveness of the system, achieving an accuracy of 99.20% in classifying job postings using the Random Forest classifier.