Decision Tree on a Real Dataset Practice Problem

This data science coding problem helps you practice Decision Trees, decision tree on a real dataset, and implementation skills. Read the problem statement, write your solution, and strengthen your understanding of Decision Trees.

• Problem ID: 205
• Problem key: 205-decision-tree-on-a-real-dataset
• URL: https://datacrack.app/solve/205-decision-tree-on-a-real-dataset
• Difficulty: medium
• Topic: Decision Trees
• Module: Supervised Learning

Problem Statement

# 🧩 Decision Tree on a Real Dataset

---

### 🎯 Goal

Train and evaluate sklearn's `DecisionTreeClassifier` on the Iris dataset.

This problem connects the earlier decision-tree concepts to a real library model.

---

### 📖 Introduction

The earlier problems showed the ingredients of decision trees:

- impurity scores
- entropy and information gain
- splitting a dataset
- choosing a good split
- traversing a trained tree

Now sklearn handles the training algorithm for us.

The Iris dataset contains flower measurements and species labels. A decision tree learns threshold questions such as:

```text
petal width (cm) <= 0.8
```

Then it uses those questions to classify new flowers.

---

### 💻 Task

Implement `iris_decision_tree_classifier(max_depth=3, test_size=0.2, random_state=42)`.

Your function should:

- Load the Iris dataset.
- Split the data into training and test sets using:
  - `test_size=test_size`
  - `random_state=random_state`
  - `stratify=y`
- Train `DecisionTreeClassifier` using:
  - `max_depth=max_depth`
  - `random_state=random_state`
- Predict on the test set.
- Compute test accuracy.
- Return model details and feature importances.

Round accuracy to 4 decimals and feature importances to 6 decimals.

---

### 📥 Input / 📤 Output

**Input**

- `max_depth`: maximum depth allowed for the decision tree
- `test_size`: proportion of the dataset used for testing
- `random_state`: seed used to make the train/test split and model training reproducible

**Output**

Return a dictionary with:

- `accuracy`: test accuracy rounded to 4 decimals
- `tree_depth`: depth of the trained tree
- `n_leaves`: number of leaf nodes in the trained tree
- `predictions`: model predictions on the test set as a list
- `feature_importances`: dictionary mapping each feature name to its importance, rounded to 6 decimals

---

### 🧩 Starter Code

```python
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score

def iris_decision_tree_classifier(max_depth=3, test_size=0.2, random_state=42):
    """
    Train and evaluate a DecisionTreeClassifier on the Iris dataset.
    """
    data = load_iris()
    X = data.data
    y = data.target

    # Your code here
    pass
```

---

### 💡 Example

```python
iris_decision_tree_classifier(max_depth=3, random_state=42)
```

Expected Output starts like:

```python
{
    "accuracy": 0.9667,
    "tree_depth": 3,
    "n_leaves": 5,
    "predictions": [...],
    "feature_importances": {...}
}
```

---

### ⚠️ Common Mistakes

- Evaluating on the training set instead of the test set.
- Forgetting `stratify=y`, which helps keep the class proportions similar in the training and test sets.
- Confusing `max_depth` with number of trees. This model trains one tree.