 I'd like to surprise someone dear with flowers. Now the surprise here wouldn't just be the flowers themselves, but also knowing their exact kind. Now I could go to a flower shop and buy irises, but irises come in various shapes and sizes. If I get them, for example, in Violet, there are three sorts – Versicolor, Virginica, and Satosa. And in taking a quick look at some pictures, I can't really tell the difference. But luckily, I'm not the first person to think about all the species of irises. Edgar Anderson collected data on 150 irises way back in 1936, and that very same dataset is available in orange. It includes three different species, described by four morphological features. More concretely, the length and the width of the sepals and petals. Now this means I could take these measurements as well, and then use them to figure out the species. There is one problem though – biology has never been my strong suit, and I don't actually know what the difference is between a petal and a sepul. After a quick google search though, it turns out it's pretty simple. So now let's say I went to the flower shop, found a bunch of irises, and measured their leaves. I would probably get some weird looks, but I'm sure they would all understand if I told them I'm a data scientist. So next step is now to type up my measurements into a table. I have the sepul length, the sepul width, the petal length, and the petal width. I have to be a little bit careful here, just to make sure I use the same names Anderson did. So these are my measurements in centimeters, and I did get a little bit distracted for the last flower, so I only measured the petals. Ok, now I can take Anderson's data set, build a model to predict the species of iris, and then use it on these three flowers that I measured. So moving on to orange, I can do this by building a classification tree to model the data. Then I can forward that to the predictions widget. And lastly I'll add my data into orange, and just quickly make sure it's all there. Ok, now that everything is set up, I can send my data to predictions, and hopefully find out what kind of irises I bought. Lastly my three flowers come from three different species. The first one is a versicolor, the second one is a sitosa, and the third one is a virginica. So let's just take another look at our workflow. We loaded Anderson's iris data set with the data set's virgin. Now notice that the data contains the species of flower, or the label, in addition to the leaf measurements of each instance. Then we used this data to build a classification tree, in other words a model that looks at leaf sizes and predicts the species of iris. Now after we have such a model, we can feed it to predictions. This widget will take a model, then use it to predict the labels of some new unlabeled data. So I put in the measurements I made, eager to find out the species of each flower. And here they are. I can also take a look at the predicted probabilities, and see that the model is very sure of itself. It's almost 100% certain of each of its predictions. Now recognizing irises is great in and of itself, but in reality my intent all along was to introduce a new chapter to our series. In the next couple of videos we'll be dealing with classification. This means we'll be answering questions like what exactly is a classification tree? How did it develop from my data? How accurate are my predictions, and can I explain them? We'll also be looking at some other modeling techniques, and see how they compare to classification trees.