Using ChatGPT to Generate and Export Data for Analysis

ChatGPT is a powerful tool that can be leveraged by data analysts that are looking to generate both real world, and mock data. In this brief demo I'll share how I've used ChatGPT to curate a table of customer purchases, quickly pivot the product the customers were purchasing, and ultimately consume this dataset in Tableau. I'll also show how ChatGPT can generate data from real world data, using US presidents as an example.

[spadhnik]

Never thought of this! Thank you for the idea.

[sammarks]

The peaks in Vermont example shows how sometimes, ChatGPT makes mistakes or shows data in a confusing way- I don't know if the data is accurate, but #4, Killington Peak, at 4, 235 ft. is taller than #2 or #3. I've found it makes mistakes in comparing other things, like the sizes of cities, or the differences between wheelchair basketball (which I play), and stand-up basketball (It claimed that for wheelchair basketball, the ball and the key are larger, which isn't true at all)

[bryanstark]

I'm using the "text-davinci-003" model and when I entered the exact same question as you in the same lowercase format without punctuation I was given generic code that looked like either json or python. I asked what language it was and it said "the code you provided is not written in any specific code." So I have no idea why I'm not getting it in a table. I even asked for the output in a table. Finally, I asked for the output as a markdown table. The table was not well formatted. So the easiest way is perhaps in CSV format or in pandas dataframe. What model are you using for this prompt?
Follow-up: I switched to davinci-002 and the result was astoundingly long. It produced a 313-line python3 script to generate random purchase data. It actually stopped with "You can run the code above with the following command:" but didn't finish. BTW my tokens are set to 2048 which is the max for davinci-002. Not what I wanted, but amazing output and perhaps very helpful to create random data sets. Actually, the function is only 46 lines. But the remaining lines are examples of how to run the code with variable data set population sizes. It says "generate_purchase_data.py -n 10" for example up to a million variable lines.

[JCEurovisionFan]

Can you do a Tableau tutorial on how to do a Likert scale-type chart?

[codygoldade]

Anyone have any tips on learning to become a data analyst? Like what would be the best way to learn? Any courses that can cover sql excel python r tableau etc?

[robertmaxey]

At this stage, before you rely upon AI generated data, you bloody well better make sure the data is accurate. I tested this last week with a small dataset and the AI was wrong. Apparently, adding numbers is still in its infancy. Had this data been used to generate other data, everything would fail.

[theoriginaljabootee]

thats wrong, the states were of the political party and not birth state