Evaluation of CNN-BiGRU and CNN-BiLSTM Model for Fake Job Post Detection: A Deep Learning Approach
Sushilkumar Chavhan, Rajesh C. Dharmik, Sachin Jain · 2024
In today's world most of the people try to search jobs from online job portals, recommendation and a very few models can predict the accurate results. This has become a big problem in the context of job data, making it more difficult for job seekers to find real opportunities. Our study uses deep learning methods such as CNN-BiGRU (Bidirectional Gated Recurrent Units) and CNN-BiLSTM (Bidirectional Long Short-Term Memory) along with, Natural Language Processing techniques like tokenization and count vectorization to address this issue. To train our models, we collected data from Google Dataset Search. Upon extensive testing and fine-tuning, we discovered that our model CNN-BiGRU and CNN-BiLSTM were highly proficient at distinguishing between authentic and fraudulent job ads. They had a high degree of reliability in identifying fake job posts and were accurate. After achieving an improved accuracy rate of +5%, we finally attained 97% accuracy in identifying and eliminating fraudulent job advertisements, and can provide job seekers with a safer and more confident online platform experience. Furthermore, our work can contribute for further developments in the area of fraud detection systems in the Digital Employment Market, Government Agencies, Social Media Platforms, Research and Policy Development.