saimaniippili/fake_job_detection — reverse-engineered prompt

Build me a simple fake job posting detection project from this repo. Use the existing notebook and the zip dataset if it has the data inside. I want to be able to train a model on job posts, clean the text, remove useless words, handle things like stemming or lemmatizing, and turn the text into features with TF IDF.

Please compare at least two models, Random Forest and SVM, then show which one performs better with clear accuracy and confusion matrix style results. Make the notebook easy to follow for someone who is not very technical, with short explanations before each major step.

At the end, add a small section where I can paste a new job description and the notebook predicts whether it looks real or fake, ideally with a confidence score. Keep everything runnable in Jupyter, fix any broken paths, and add helpful comments so I can present it as a complete machine learning project.