Your First Machine Learning Application: Classifying Iris Species

Before you write any machine learning code, you need the right tools installed. The tool used in this material is scikit-learn, a Python library for machine learning. Different releases of scikit-learn do not always offer the same modules or the same import paths. A program that runs cleanly on one release can fail on an older one, because a function or module it depends on was not yet available when that older release was published.

This article uses the version requirement that comes with your first real application: classifying iris species. The requirement is short, but it teaches a useful lesson about how library versions connect to the code you write. Once you see how that connection works, you can check any version requirement you meet in the same way.

The application is built around a hobby botanist who has found some iris flowers and wants to tell their species apart. For every iris she has measured four quantities in centimeters: the length and the width of the petals, and the length and the width of the sepals. Those four numbers describe one flower. Together they form a single record, and the whole problem is built out of many records like it.

Keep this one flower in mind as you read. Her four measurements tell you what the flower looks like. They do not, by themselves, say what species it is. Separating those two roles is the central move of this article.

The botanist's situation is a supervised learning problem. In supervised learning, we have measurements for which we already know the correct output. The known correct outputs are what set this setting apart: the model is given examples together with their answers, and it learns the relationship between them.

The four measurements are the inputs: the features that describe each flower. The species name attached to that same flower is the label. When the model is trained, both pieces are visible together. When it is later asked to predict, the label is the thing it has to supply from the measurements alone. The diagram shows this split for a single record: four input cells that describe the flower, and one label cell that names it.

Within supervised learning, this application is a classification problem. In a classification problem we want to predict one of several options, and here those options are the species of iris. The possible outputs are called classes.

Every iris in the dataset belongs to one of three classes: setosa, versicolor, or virginica. Because there are exactly three possible outputs, the problem is described as a three-class classification problem. The number of classes is a property of the problem itself, and it tells you how many distinct answers the model is allowed to give.

The two terms are easy to mix up because they describe the same problem from different angles. Supervised learning describes how the model is trained. Classification describes what kind of output is wanted. The iris application is both at once: it is trained on labeled examples, and its output is one class out of three.

The material assumes that you have version 0.18 or later of scikit-learn. Read that as a floor, not a target. Anything at 0.18 or above is what the material expects, and the material also recommends a version at least as recent as the one its authors used.

The reason 0.18 is the floor is the model_selection module. That module was added in version 0.18. Any code that imports from model_selection depends on the release that contains it. If you install an earlier version, imports from this module will not resolve as written, and you will need to adjust them to match what your installed version provides.

A version check is a quick way to confirm that your environment meets the floor before you run any iris code. The first step is to import scikit-learn and print the version string it reports. The second step is to confirm that the model_selection module can be imported. If the second import fails, your installed release is older than the one this material assumes.

Reading the output works like this: if the version printed is 0.18 or higher and the second line runs, your environment meets the requirement. If the import raises an error, the release is older than 0.18, and the imports from model_selection will need to be adjusted to match what that release provides.