Master Decision Trees with the Iris Dataset: A Hands‑On Guide

1. Classification and Decision Tree Introduction

Classification is the process of discovering patterns in data to make judgments, such as email spam filters that learn from user‑labeled messages.

The typical classification workflow includes: (1) labeling training data as positive or negative, (2) training a model, and (3) using the model to predict and evaluate new data.

Decision Tree is a widely used technique for classification and prediction. It builds a tree‑structured model that relates attributes to class labels, allowing fast inference. Common algorithms include CART, ID3, C4.5, CHAID, Decision Stump, Random Forest, MARS, and Gradient Boosting Machine.

Examples illustrate how decision trees mimic everyday decision making, such as selecting a partner based on age, appearance, income, and occupation.

2. Iris Flower Dataset

The Iris dataset, included in scikit‑learn, contains 150 samples of three Iris species (setosa, versicolor, virginica) with four numeric features: sepal length, sepal width, petal length, and petal width.

The dataset provides two main arrays: iris.data (the feature matrix) and iris.target (the class labels).

Typical output shows the first 50 targets as 0 (setosa), the next 50 as 1 (versicolor), and the last 50 as 2 (virginica).

3. Decision Tree Implementation on the Iris Dataset

DecisionTreeClassifier from scikit‑learn implements a multi‑class decision‑tree classifier.

Key parameters are the feature matrix X (shape [n_samples, n_features]) and the label vector y (shape [n_samples]).

Sample code to train and predict on the Iris data:

The resulting classification separates the three Iris species.

Code Optimizations

Two issues were addressed: (1) using all four features instead of only the first two, and (2) splitting the dataset into training (70%) and testing (30%) subsets.

Training‑test split example:

Full optimized code (including accuracy and recall metrics):

Sample output shows classification performance metrics.

Visualization of the decision tree:

Additional Knowledge

An example from the scikit‑learn documentation demonstrates another decision‑tree workflow.

When visualizing the tree with Graphviz, errors may occur; the article shows the iris.dot content and the command‑line steps to generate the image.