Baselines Matter: Understanding DummyClassifier and “Most Frequent” Strategy

-

 # Baselines Matter: Understanding DummyClassifier and “Most Frequent” Strategy


When you’re starting a machine learning project, a strong baseline is your best friend. It tells you whether your fancy model is actually learning anything useful—or just adding complexity. In scikit-learn, DummyClassifier is the go-to tool for building such baselines.


This post walks through a concrete example, explains why the result looks the way it does, and shows how to use baselines responsibly.


---


## The Setup


- Feature matrix: X with shape (1000, 5)

- Labels: y with two classes {0, 1}

- Class distribution: 650 examples are class 1, 350 are class 0

- Code:


```python

from sklearn.dummy import DummyClassifier


base_clf = DummyClassifier(strategy='most_frequent')

base_clf.fit(X, y)

print(base_clf.score(X, y))

```


What happens here?


- `strategy='most_frequent'` always predicts the majority class seen during `fit`.

- Since class 1 appears 650/1000 times, the classifier will always predict 1.


Therefore, the training accuracy printed is:


- Accuracy = correct predictions / total = 650 / 1000 = 0.65


So, the output is 0.65.


---


## Why This Baseline Is Important


- It quantifies the minimum performance a model must beat. If your logistic regression or random forest scores around 0.66 on the same data, it’s barely better than predicting the majority class every time.

- It highlights class imbalance. A seemingly “good” accuracy (65%) is actually trivial to achieve here and may be useless for minority-class detection.


---


## Beyond Accuracy: Metrics That Matter


For imbalanced data, accuracy can be misleading. Prefer:


- Precision, Recall, F1 (especially for the minority class)

- ROC AUC and PR AUC (Precision-Recall AUC is often more informative with imbalance)

- Confusion matrix to see error types


```python

from sklearn.metrics import classification_report, confusion_matrix


y_pred = base_clf.predict(X)

print(confusion_matrix(y, y_pred))

print(classification_report(y, y_pred, digits=3))

```


With the most-frequent strategy (always predicting 1):

- True Positives: 650

- False Positives: 350

- True Negatives: 0

- False Negatives: 0

- Recall for class 1 = 1.00, but precision for class 1 = 650 / (650+350) = 0.65

- Class 0 is never detected (recall = 0.00)


---


## Better Baselines You Should Try


DummyClassifier supports multiple strategies:


- `most_frequent`: always predict the majority class

- `stratified`: predict according to class distribution

- `uniform`: predict uniformly at random

- `constant`: predict a user-specified label


```python

DummyClassifier(strategy='stratified', random_state=42)

DummyClassifier(strategy='uniform', random_state=42)

DummyClassifier(strategy='constant', constant=0)

```


These help you understand whether a “real” model beats random or distribution-aware guessing.


---


## Train/Test Protocol: Avoid Self-Congratulation


The code above scores on the training set. For meaningful evaluation:


```python

from sklearn.model_selection import train_test_split


X_tr, X_te, y_tr, y_te = train_test_split(

    X, y, test_size=0.2, stratify=y, random_state=42

)


base = DummyClassifier(strategy='most_frequent')

base.fit(X_tr, y_tr)

print('Test accuracy:', base.score(X_te, y_te))

```


- Use `stratify=y` to preserve class ratios across splits.

- Compare every real model against this test-set baseline.


---


## Key Takeaways


- The “most frequent” baseline here yields 0.65 accuracy because 65% of labels are class 1.

- Always create and report a baseline—then ensure your model meaningfully exceeds it.

- Don’t rely on accuracy alone; use recall, precision, F1, and PR AUC for imbalanced data.

- Evaluate on a stratified train/test split to avoid misleading conclusions.


If you’d like, share your dataset goal (e.g., maximize recall for class 0), and I can suggest tailored metrics, thresholds, and next-step models.

Comments

Post a Comment

Popular posts from this blog

Understanding Data Leakage in Machine Learning: Causes, Examples, and Prevention

-
When building machine learning models, achieving high accuracy on your training or validation data is exciting — but sometimes, that success can be misleading. One common trap that causes overly optimistic results is data leakage . In this blog, we will explore: What is data leakage? Why is it harmful? Real-world examples to understand leakage better How to prevent data leakage in your projects What is Data Leakage? Data leakage happens when your machine learning model has access to information during training that it wouldn’t realistically have when making predictions in the real world. This extra information “leaks” from the future or from the test set, causing the model to cheat. When leakage occurs, your model's performance metrics (like accuracy, R², or F1-score ) look great, but this performance won’t generalize to new, unseen data. Why is Data Leakage a Problem? Overfitting to training data : The model learns shortcuts rather than meaningful pattern...
Read more

🌳 Understanding Maximum Leaf Nodes in Decision Trees (Scikit-Learn)

-
When working with decision trees in machine learning , one of the most common questions is: 👉 How many leaf nodes can a decision tree have, given certain constraints? Let’s walk through a real example. 📌 The Question We initialize a decision tree as follows: from sklearn.tree import DecisionTreeClassifier clf = DecisionTreeClassifier( max_depth =4, random_state =42) We’re told: The dataset has 100 samples and 10 features . The tree is a perfectly balanced binary tree . No pruning is applied. No additional stopping criteria (like min_samples_split , min_samples_leaf , or max_leaf_nodes ). 👉 What is the maximum possible number of leaf nodes in the decision tree? 🌲 Key Concepts 1. Depth vs. Levels A binary tree with max_depth = d has at most d splits from the root to a leaf. Depth d = 4 means there can be 4 splits from the root node down to a leaf node. 2. Maximum Leaf Nodes Formula In a perfectly balanced binary tree : Number of leaf n...
Read more

Linear Regression with and without Intercept: Explained Simply

-
If you’re new to machine learning, linear regression is one of the easiest models to understand. It tries to draw a straight line (or a flat plane, if there are more features) that best fits the data. When we use scikit-learn’s LinearRegression , we can decide whether the line should include an intercept (a constant number added at the end of the equation) or not. The Example Code import numpy as np from sklearn.linear_model import LinearRegression X = np.array([[12, 13], [23, 24], [24, 25], [35, 36], [36, 37], [37, 38]]) # Create target values y y = np.dot(X, np.array([4, 5])) + 6 reg1 = LinearRegression(fit_intercept=True).fit(X, y) s1 = reg1.score(X, y) reg2 = LinearRegression(fit_intercept=False).fit(X, y) s2 = reg2.score(X, y) What’s Happening Here? X is our input data (two features per row). y is the output we want to predict. We made it using this formula: y = 4 ∗ x 0 + 5 ∗ x 1 + 6 y = 4 * x_0 + 5 * x_1 + 6 This formula clearly has an intercept = 6 . Case 1: Wi...
Read more