Introduction
This article presents a comprehensive walk-through of using the Code Interpreter feature within ChatGPT to analyse a publicly available dataset, generate visualizations, and construct a predictive model, without requiring any prior experience in programming (Python or any other language). The dataset we’ll be utilizing in this guide is the “Heart Disease” dataset, sourced from the UC Irvine Machine Learning Repository. The dataset in question comprises data related to heart disease patients and healthy individuals, encapsulated in multiple files of diverse formats. Detailed manual inspection of these files isn’t necessary, as ChatGPT has the ability to independently identify the content of the archive, comprehend the data formats, and discern the implied information.
While an understanding of programming is not mandatory for this exercise, proficiency in interacting with ChatGPT—composing effective prompts and subsequent inquiries—is advantageous. Also, familiarity with data science terminology could be beneficial in the current scenario.
This guide aims to provide a well-structured sequence of prompts that can facilitate productive interaction with ChatGPT. Through minimal user input, we’ll demonstrate how to efficiently clean data, draw meaningful conclusions, create visualizations for enhanced comprehension, and build a prediction model.
Enabling Code Interpreter
Please be aware that while the Code Interpreter is an innovative addition to the capabilities of ChatGPT Plus, it is not automatically activated. To utilize this feature, you must first be a subscriber to ChatGPT Plus. Once subscribed, follow the steps provided below to enable the Code Interpreter feature.
In the Setting menu, click on Beta features.
- Click on the Code Interpreter toggle. You don’t need to activate the plugin beta feature or go through the ChatGPT plugin store.
- Go to create a new prompt and hover over GPT-4 at the top. …
- Click on the Code Interpreter option.
- The Code Interpreter plugin mode is now enabled!
Conversation with Code Interpreter
See the picture below for the whole conversation with Code Interpreter. For the original conversation, follow this link to the ChatGPT website. This will allow you to copy the prompts and see the python scripts generated by Code Interpreter (but unfortunately not the images).
Conclusion
To wrap up, in this article we demonstrated how to use Code Interpreter to analyse a dataset, perform exploratory data analysis, formulate observations, create visualizations and train a prediction model, with minimal prompts for the user. ChatGPT also provided Python code which can be copied or downloaded and used to reproduce the analysis. Don’t forget to follow this link to see the conversation on the ChatGPT website.
