How soum-cs/fake-job-internship-prediction works
This repository is a research project for detecting fraudulent job postings from posting text and screenshots. The main workflow uses the shared dataset in two experiments: Experiment 1 establishes classical and multimodal baselines; Experiment 2 uses a pretrained ResNet-50, validation-based model selection and threshold tuning, then predicts genuine versus fake. The notebooks combine data preparation, modeling and evaluation rather than exposing separate runtime services. The report documents both experiments and their results. The README is empty; relationships below are limited to those stated in the report or shown in the Experiment 2 builder excerpt.
Main entry points: Experiment_2_Multimodal_ResNet50.ipynb ResNet experiment · Experiment_1.ipynb Baseline experiment · fake_job_postings.csv Job postings · images Posting screenshots
Components 11
Research data
- Job postings
Dataset/fake_job_postings.csv - Posting screenshots
Dataset/images
Detection experiments
- Baseline experiment
Experiment_1/Experiment_1.ipynb - ResNet experiment
Experiment_2/Experiment_2_Multimodal_ResNet50.ipynb - Text features
Experiment_2/Experiment_2_Multimodal_ResNet50.ipynb - Image model
Experiment_2/Experiment_2_Multimodal_ResNet50.ipynb - Multimodal fusion
Experiment_2/Experiment_2_Multimodal_ResNet50.ipynb - Training and selection
Experiment_2/Experiment_2_Multimodal_ResNet50.ipynb - Posting prediction
Experiment_2/Experiment_2_Multimodal_ResNet50.ipynb
Results and reporting
- Experiment reports
Docs/Fake_Job_Fraud_Detection_Report.docx - Experiment comparison
Docs/Fake_Job_Fraud_Detection_Experiments_Report.docx
Connections 11
- Job postings → Baseline experiment: uses corpus
- Posting screenshots → Baseline experiment: uses screenshots
- Job postings → ResNet experiment: uses corpus
- Posting screenshots → ResNet experiment: uses screenshots
- ResNet experiment → Text features: extracts text
- ResNet experiment → Image model: processes images
- ResNet experiment → Multimodal fusion: combines branches
- ResNet experiment → Training and selection: trains and selects
- ResNet experiment → Posting prediction: predicts postings
- Baseline experiment → Experiment reports: documented in
- ResNet experiment → Experiment comparison: compared in