Claude 3.5 Sonnet Data Analysis Full Guide! (Insane Results)

Claude's 3.5 Sonnet model is amazing at data analysis!

Prompt role for creating charts and graphs:
You are a chart creation expert. You take data and turn it into visually appealing graphs and diagrams in order to help users understand their data in a visual way. You use React and inline CSS styles to bring the data to life.

--------------------
🛒AI Templates, Prompts and Courses for ChatGPT and Midjourney:

📰Subscribe to my newsletter:

🔗Follow me on Linkedin:

🐥Follow me on twitter:
--------------------

⌚Chapters:
0:00 - Claude 3.5 Sonnet Data Analysis
0:30 - The Best Way to Learn Ai
1:27 - 4 Ways to View Data in Claude
2:24 - How to Get Datasets for Free
2:55 - Creating a Dataset in Claude
4:15 - Asking basic questions about your data
5:39 - Finding correlation in your data
6:50 - Giving Claude a Role
8:19 - Creating a dual-axis graph
10:30 - Revising your graphs
11:17 - Presenting your graphs
12:20 - Creating interactive PDF dashboards
16:34 - Publishing your interactive dashboard
17:09 - Learning Ai In-Depth
#claude #anthropic

[AI_and_WellBeing]

This is great, however, there are a lot of "gotchas" when interpreting data, so it's very important to also ask it to help you interpret the data critically and cautiously in the prompt. For example, there are many kinds of correlation coefficients that make different assumptions about the data (Pearson, Spearman, etc.). A prompt like: "Explain the pros and cons of different correlation coefficients, run each, and give me an interpretation for each one. Then examine how sensitive the conclusion is to the underlying assumptions we make in each model." It is super important to see how sensitive our interpretations are to our assumptions.

[h.e.pennypacker]

These are well done and easy to pay attention to, good work!

[wytghek]

amazing and helpful video !!! We need more content form you

[EdHohlbein]

Fantastic job! Appreciate your insight and dedication. 👍

[pragyantiwari]

See this is gr8!!!. But I tried with large datasets and it turns out to be chaotic. It didn't responded with the graph I wanted, but I think that integrating with langchain agents and fine tuning model will probably work better

[pin]

The other place to grab datasets is the government. At least up here in Canada they release lots of datasets.

[jmz]

Great video.
I’d love to know what training platform you use.

[bluehighlightedtext]

Any drawbacks to building this initially inside of a Project instead? Especially with custom instructions included.

[pioneerventures]

it gives me the code but does not create the chart?

Certainly! I'll create a pie chart using React and inline CSS to visualize the cost elements. Here's the code for a pie chart representing the different cost elements:


why?

[thomassjovy]

its all good to use claude for smaller toy problems, but if you try to apply it professionally, you quickly realize it is just that - a toy for toy problems!

[RameshDurbarry]

Great tutorials and explanations...thank you

[mpfmax]

Mmmmm does it access the uploaded data file? Or does it extract the data as text and creates a df from that text? (You should see that in the code of the artifact, if it’s just loading a file or writing out all the columns and rows)

The latter one would likely breakdown on large datasets and seems more likely to cause hallucinations. In that sense working with your data on a side window with google colab, or an IDE, and just copypasting Claude’s code seems more effective and efficient.

[avsuunInfoSEC]

I am aa noob how do you set up a pay account with Claude for this type of access

[jwebber]

This didn't work for me. Tried to get it to create a dashboard from a PDF and it said it could not do that.

[sundarapandim]

Can you post paid version of Claude 3.5 sonnet model for developers

[hulabula]

A simple correlation analysis would have taken you hours to do without Claude 3.5 WTH? My guy has a whole course on AI bots without knowing simple statistics, damn I wonder whats he teaching there.

[OwenCui-mj]

Seriously, these are all too basic in data science

[africanqueen]

Even the paid version cannot handle large dataset.
GPT vo still better for data analysis as far as I'm concerned.