So I have been using ChatGPT quite a lot in the past six months and I have always wondered if I can do a full-blown machine learning project just using ChatGPT. I found a really cool dataset consisting of phishing emails and safe emails, and I want to see how ChatGPT approaches this task. To start off, I created a Jupyter notebook and imported some basic libraries.
Next, I explored the dataset and noticed that it contains email text and email type. I decided to use the CountVectorizer to represent the pre-processed email text and applied regular expressions for text cleaning. I also converted the email text into a list of values.
After the pre-processing step, I used logistic regression to classify the emails. Surprisingly, the model achieved an accuracy of 97%. I then tried other machine learning models such as random forest, support vector machine, and gradient boost. However, these models took longer to run and achieved lower accuracies.
In conclusion, I have successfully explored machine learning using ChatGPT and achieved high accuracy in classifying phishing emails. Logistic regression proved to be the most efficient model for this task. I am excited to continue exploring the capabilities of ChatGPT in future projects.
