Fraudulent Job Posting Detection Using Logistic Regression
Vijay Madaan, Neha Sharma, Raghubeer Singh Bangari, Srinivas Aluvala · 2024
In the present study, the application of logistic regression models to analyze language technologically driven illegal job advertising identification is explored. With a whole collection of linguistic characteristics—word rates and n-grams derived from job ads, we built a scenario that uses a logistic regression model to identify ads as either genuine or fake. A remarkable accuracy of 98.19% on the test set was obtained, confirming the subject's resolve in distinguishing between dishonest and honest job advertising. An analysis was conducted to determine the effect of several critical factors, including modulation amplitude and optimal technique, on the model's performance. To achieve a harmonious equilibrium between the model's accuracy and its capability to operate effectively on unfamiliar data, modifications were implemented to these hyperparameters. Furthermore, an analysis was conducted on the framework's loss function to obtain precise insights into its convergence characteristics during the training stage. Our results highlight the need of logistic regression methods in spotting phony job postings and clarify how hyperparameter modification affects model performance. This work contributes significantly to the corpus of knowledge on recognizing frauds in job site promotion and presents vital new insights on the appropriate use of logistic regression analyses for this objective. The report also identifies areas that could benefit from additional investigation, such as comparing various machine learning approaches and introducing more text components, to enhance the possibility of identifying falsehoods.