Logistic regression on the Iris data set

Mon, Feb 29, 2016

The Iris data set has four features for Iris flower.

• sepal length
• sepal width
• petal length
• petal width

Using a three class logistic regression the four features can be used to classify the flowers into three species (Iris setosa, Iris virginica, Iris versicolor).

Using this Jupyter notebook combinations of two features we are used to classify the species. The mis-predicted values are shown below.

measure 1	measure 2	incorrect predictions
sepal length	sepal width	29
sepal length	petal length	6
sepal length	petal width	8
sepal width	petal length	7
sepal width	petal width	7
petal length	petal width	6

The previous post shows that some combinations of features are easier to use to separate the species than others.

Logistic regression can also be used on the two principal components and mis-predicts five specimens.

A mesh when drawn over the plot shows the three classes of the logistic regression.