CIS4301 Chapter12 Part3

Añadido: 2026-05-09

[00:00:00]Okay, we're on page 617. We just finished talking about supervised versus unsupervised uh data mining. So, unsupervised doesn't require any smarts whatsoever. You just turn the machine loose. Now, I'll confess, you need a little bit of smarts to figure out what it says, but to actually do the data mining step, nothing. Okay. On the other hand, um if I'm trying to figure out exactly one thing, you know, then the supervised one takes lots and lots of skills. So they talk about the skills involved. You need people that are good in math [laughter] and artificial intelligence and data and marketing. And so for example, if I'm trying to do a supervised look at my data to find figure out, should I have yet another model of lawnmowers? I mean, I already have three models of lawnmowers, but I think maybe a, you know, a high-end model might actually be good. So what's the market for a high-end lawnmower? Well, trust me, you're going to need some very smart people who know the business of lawn care. Not data management people unless they mow their own lawn, right? Not artificial intelligence, not mathematicians. You're going to need people who actually know about marketing lawnmowers, right? Okay, good. Okay. So, there's all the point of this graphic is there's an awful lot of skills that are involved in this thing and you can't necessarily have one person do them all.

[00:01:25]So when you're doing a supervised, you need lots and lots of people. Okay. So they talk about popular techniques on page 618. So um one is called uh a market basket analysis. So here here is remember this dive shop where they have mass fins and weights etc. And um they're trying to figure out the likelihood of support confidence and lift. I'm not going to go into all there. Basically what it talks about if if you bought uh you know if you bought fins and masks and so what they're trying to do is they're trying to predict you know what you're going to buy next so to speak u so based upon what you have purchased in the past what you might so a good example is I go to Amazon and I buy a book on car sharp language and it says oh this is an easy one just says oh wow here's 14 other books on the topic of C. I'll offer him that. That's an easy one that hardly takes any effort whatsoever to pull that one off. So, market basket analysis.

[00:02:32]What have you been buying? So, therefore, what should I stock in the warehouse next? Okay. Decision tree is more about personal things. This is the big picture. This is what should I put in the warehouse? I'm looking at what people are buying and so I need to know what to put in the warehouse. Decision tree is more along the lines of well this is kind of a goofy example but you know it's one of those things where I'm trying to predict one person okay what are you going to buy next like the example of buying a book in Amazon there's one called logistics regression and it's like this one is kind of a cool one they have an example there some of this is is just kind of easy to get your head around what is the likelihood of you going ice fishing well probably has a lot to do with the weather. Okay, if you're going ice fishing, I don't want to be out there when it's, you know, 20 below and in and a 40 mph wind, right?

[00:03:29]So, figuring out the likelihood of you going ice fishing based upon the weather, that's logistical regression.

[00:03:35]And then neural networks. Neural network is for solving things where the math is not well understood. You know, there's no real formula. All I know is this sometimes is correlated to that. Maybe not all the time. So maybe there's a correlation between this and this, but there's something else going on and I don't know what it is. So I can't predict in advance. So neural networking is all about trying to make sense of things where there are patterns that are disruptive patterns. Typically, it's where the the very complex math or perhaps even where the math is not well understood. That's what neural networks are all about.