ONLINE JOB SCAM DETECTION USING SMART SYSTEM

Abstract

In light of the growing world breaches of data that occur on a daily basis. With every successive day, the number of job seekers who fall prey to a becomes dramatically more and larger. The bulk of jobseekers are recruited by corporations and fraudsters use the approaches, with the majority of them coming from digital job-posting in Indeed website. We want to apply Machine Learning in the future to reduce the incidence of similar frauds. Candidates would be able to maintain their vigilance and make smart decisions, when necessary, hence lowering the amount of such frauds that take place in the first place. Natural language processing will be sometimes used investigate the attitudes and patterns in the job advertisement, and to do so (NLP). Later, models like Logistic Regression, Naïve Bayes and some other classification algorithms were used to classify the data into fraudulent or non fraudulent job posting. Recently, some people have also used regular artificial neural networks and LSTM along with Bert or 200 dimension Glove for word embedding. All these models have achieved very good accuracies.We have implemented Logistic regression and Naïve Bayes as our baseline models to see how they perform on our unbalanced data.To make the data more balanced, we oversampled it and used it to test our baseline models.We noted that the model has improved its performance . To improve the accuracy we have chosen to work with CNN models on our oversampled data and used Google's pre-trained model Glove 300d to perform word embedding. We have evaluated all our models and found out that CNN model works the best by achieving the highest accuracy.