How AI works in data analysis

[jahid12]

For those of you who have never tried using these AIs canada phone number list for analysis (because we understand that using them to ask questions or write questions has been something you've probably done before), we'll explain how it works.

We start by using these AI-based chats (ChatGpt, Claude, Gemini) as we always do: writing and requesting what we're looking for. How we write it, provide context, and guide the machine is an art and would be the subject of another post. But even without much practice, just asking will get us results.
We ask AI for things, as always, but in this case, we'll do so with a direct focus on analysis, obtaining results and conclusions from the data we have available. To do this, we'll need to get that data onto our computer (usually in CSV format) and upload it to the AI ​​so it can process it. This can also be done by uploading images, but it's usually much less effective because we're forcing it to interpret them.
This analysis is automatic in ChatGPT and Gemini; most of their models can perform it without problems (and in the models that can't, it simply can't be performed). In Claude, however, we must have this enabled. We'll need to go to our user's configuration options and activate the "Analysis Tool" feature (under Features).
This system works very similarly to all AIs: 1. The system will tell us that it's "Analyzing" and, behind the scenes, it will write the programming code with which it will process the data to classify and optimize it. You can see that code by clicking on it, but it's hidden by default because it's assumed you're not interested. 2. Then, the AI ​​will execute that code on its own, retaining the result, and then be able to continue working. If you wish, and understand some programming, you can open the generated code to see the steps it's taking to analyze the content.
After this initial analysis, the AI ​​returns to its generative mode and can now provide us with insights into what it has extracted, but it doesn't end there.
Another (optional) detail is the ability to show us data visualizations or return pre-processed data. Each AI takes a different approach to this, and they differ slightly, but the goal is the same: to communicate the results to us through text, visualizations, or documents.
Thus, when used properly, the system can perform in just a few minutes what would normally take hours or full days of work.
This feature, which sounds so good when explained, is very useful in practice, but far from perfect. Some analyses aren't performed well, and some visualizations are... let's say... difficult to visualize. But that doesn't mean it's not a tool we should master and incorporate into our work. To that end, I spent a few days testing different systems to draw conclusions: Which one should I choose? Well, that's what this post is about...