Editor's Image
ChatGPT has become the OpenAI product that changes the way the world works. Many of the readers here are already using them or at least trying them. The way it helps us, I don't think we can go back to our old way of working.
One of the innovations that OpenAI offers is the GPT Store, where people can develop their custom GPT models and share them with the public. More than 3 million ChatGPT custom GPTs from creators are open. In fact, some of it could be useful for improving the work of data scientists.
This article will discuss 7 GPTs from GPT Store that could improve your data science workflow. What are these GPTs? Let's get into it.
As a side note, I would use the Kaggle Telecom Churn Dataset as an example data set for GPTs to use.
Let's start with what the ChatGPT teams have created for us, the Data analyst. This is the custom GPT, explicitly trained to analyze our data and visualize it as we need. By dropping the file, such as CSV files, and providing the message of what we need, Data Analyst GPT could do the job automatically.
For example, I ask the data analyst to develop a churn correlation analysis based on the data set I provide.
Data analyst performs correlation analysis (Image by author)
You can request further analysis from the data analyst GPT. You can also use GPT to provide the complete code and run it yourself if necessary.
The next GPT we will discuss is the GPT machine learning. This custom GPT is designed as an assistant for any data science and machine learning activities. Utility includes discussing, learning and developing algorithms suitable for our data projects.
As an example, I ask the machine learning GPT to perform model development from our example data set to predict churn. Here is the result.
Machine Learning performs model experiments (Image by author)
The GPT can provide an excellent comparison between the models they used. If we continue, we can ask the model to iterate with more models, perform hyperparameter tuning, and ask the GPT to provide reasons for each action.
As in the previous entry, the GPT Machine Learning Engineer provides users with a wizard to develop the machine learning model. You can put in your dataset and ask GPT to provide you with the essential steps and complete code.
What sets the machine learning engineer apart is that their GPT specifies model design to automate complex tasks, especially for model deployment. The GPT is good for discussing how you want to structure your model and what the deployment of the model to production should look like.
Speaking of model structuring, GPT is also suitable for helping us structure our code for machine learning modeling. One of the best coding assistants I have found is the AutoExpert. It is a GPT that is designed to help you as a constant pair programming assistant.
The GPT is built with additional code generation capability, online access to the latest APIs, and custom commands to save your session state, which you can use for a later session if needed.
Using this GPT will help you generate the complex code for any purpose you need during data science activity. It also provides you with the code and script structure to help you execute them better.
Let's move on from the technical coding part and go to the theoretical one. As we know, data science jobs are about continuous learning, especially in novel use cases. With the increasing research in data science, sometimes it is difficult to find the perfect research that can help in our use cases. This is where AcademicGPT comes in.
This GPT will help you find the latest research works for our use cases. From a simple message, it would give us a selection of the latest articles related to the problem we want to solve.
For example, the text below is the output of ScholarGPT, where I uploaded our dataset and asked them to provide me with a research paper related to attrition prediction.
Title: “Transparency in decision making: the role of explainable ai (XAI) in customer churn analysis”
- Authors: C ÖZKURT
- Year: 2024
- Summary: This study focuses on predicting customer churn and explaining the reasons behind it using machine learning, specifically looking at customer churn in the telecom sector through rigorous analysis.
- Link: Read the paper?fountain?.
ScholarGPT provides many more research articles for you to choose from, so you can select which one applies to your use cases.
The next GPT we would discuss is the Whimsical diagram. For many data science activities, it is not always about research and model development. There are many times that we need to visualize our workflow and give an explanation of what our work would be like. This is where Whimsical Diagrams GPT will help you.
This GPT is designed to explain and visualize concepts with flowcharts, mind maps, and sequence diagrams. Providing the source of data and indications we have could help us provide a visualization that would help our work.
For example, I asked the model to provide me with a hint diagram of the churn dataset and it suggested visualizing churn by features. Below is the image result.
Rotation by features (Image generated by Whimsical Diagram GPT)
You can continue talking to GPT to find the perfect diagram workflow for data science jobs.
The last one is the CanvaGPT, which could help us communicate our results. As we know, Canva is a service platform that helps design everything from logos to profile photos, banners, and presentations. With Canva GPT, they can help us get the best design for our analysis.
Data science is about communicating our results to others, so it is essential to have valid results that are presented in a way that the audience understands. With Canva GPT, we can ask for suggestions on which design is appropriate. For example, I asked the model to provide me with a layout that would be perfect for presenting churn statistics.
Churn Statistics Layout Selection (Canva GPT)
The GPT would give us design options and we could choose which one we preferred or give additional indications to obtain other designs.
This article discusses seven custom GPTs available in the GPT Store that could improve our data science workflow; they are:
- Data Analyst by ChatGPT
- Machine Learning by Maryam Eskandari
- Hustle Playground Machine Learning Engineer
- AutoExpert (developer) by llmimagineers.com
- ScholarGPT by awesomegpts.ai
- Whimsical Diagrams from whimsical.com
- Canva by canva.com
I hope that helps! Do you have any GPT suggestions that should be on this list? Leave them in the comments too.
Cornellius Yudha Wijaya He is an assistant data science manager and data writer. While working full-time at Allianz Indonesia, she loves sharing Python tips and data through social media and print media.