Introduction to Machine Learning: One-Day Course

This is a one-day machine learning introductory course for beginners. The course covers the basics of supervised and unsupervised learning, including regression, classification, clustering, dimensinality reduction and anomaly detection. It also includes hands-on exercises and examples using popular Machine Learning (ML) libraries like Scikit-learn.

The presentation is used to guide the instructor through the course, providing a structured outline of the topics to be covered.

Image created by using ChatGPT

Table of Contents

• Introduction to Machine Learning: One-Day Course
  • Table of Contents
  • 1. Introduction to Machine Learning
  • 2. Understanding the Machine Learning Workflow
    • Steps in the Workflow
  • 3. Supervised Learning
    • 3.1 Regression
    • 3.2 Classification
  • 4. Unsupervised Learning
    • 4.1 Clustering
    • 4.2 Other Unsupervised Learning Techniques
  • 5. In-Class Assignment
• Using with GitHub Codespaces
• Acknowledgements

---

1. Introduction to Machine Learning

• What is AI and ML?
  • AI: The ability of machines to simulate intelligent behavior.
  • ML: A branch of AI where models are trained to learn from data and improve over time.
• Applications:
  • ChatGPT, Netflix recommendations, fraud detection, self-driving cars, etc.
• Types of ML:
  • Supervised Learning, Unsupervised Learning, Reinforcement Learning.
• Key Terminology:
  • Dataset, Features, Labels, Model, Training, Testing, Hyperparameters, Overfitting, Underfitting.

Highlights

• Comparison between supervised and unsupervised learning using Linear Regression and K-Means examples.
• Basic visualizations of regression and clustering tasks.

Related notebook: Introduction to Machine Learning

---

2. Understanding the Machine Learning Workflow

Steps in the Workflow

1. Define the Problem: Set objectives (e.g., regression, classification).
2. Collect and Clean Data: Handle missing values, duplicates, outliers.
3. Explore and Visualize Data: Use summary statistics and visual tools like histograms and scatterplots.
4. Feature Engineering:
   • Selection: Remove irrelevant features.
   • Transformation: Normalize and encode data.
   • Creation: Generate new features.
5. Split Data: Divide into training, validation, and test sets.
6. Choose and Train a Model: Select an algorithm based on the task.
7. Evaluate the Model: Use metrics like RMSE, Accuracy, and Silhouette Score.
8. Hyperparameter Optimization: Use GridSearchCV or RandomizedSearchCV for fine-tuning.

Highlights

• End-to-end example of an ML pipeline using Scikit-learn.
• Visualization of preprocessing and evaluation results.

Related notebook: Machine Learning Workflow

---

3. Supervised Learning

3.1 Regression

• Goal: Predict continuous outputs (e.g., house prices, temperature).
• Common Algorithms:
  • Linear Regression, Polynomial Regression, Ridge, and Lasso.
• Evaluation Metrics:
  • MAE, MSE, RMSE, ( R^2 ).

Highlights

• Hands-on example of Linear Regression with visualization of results.
• Analysis of regression coefficients.

Related notebook: Supervised Learning - Regression

---

3.2 Classification

• Goal: Predict discrete categories (e.g., spam detection, disease diagnosis).
• Types:
  • Binary, Multi-Class, Multi-Label Classification.
• Evaluation Metrics:
  • Accuracy, Precision, Recall, F1-Score, Confusion Matrix.

Highlights

• Logistic Regression example for binary classification.
• Hands-on exercise with Random Forest Classifier.
• Visualization of confusion matrix results.

Related notebook: Supervised Learning - Classification

---

4. Unsupervised Learning

4.1 Clustering

• Goal: Group data points into clusters based on similarity without labels.
• Types:
  • Partition-Based: K-Means.
  • Hierarchical: Agglomerative Clustering.
  • Density-Based: DBSCAN.
• Evaluation Metrics:
  • Silhouette Score, Inertia, Visualization.

Highlights

• K-Means Clustering example with synthetic data.
• Visualizing clusters and centroids.

Related notebook: Unsupervised Learning - Clustering

---

4.2 Other Unsupervised Learning Techniques

• Dimensionality Reduction:
  • Reduces input features while preserving patterns.
  • Example: PCA (Principal Component Analysis).
• Anomaly Detection:
  • Identifies outliers or unusual patterns.
  • Examples: Isolation Forest, Z-scores.

Highlights

• PCA visualization of high-dimensional data projected into 2D.
• Hands-on example of Isolation Forest for anomaly detection.

Related notebook: Unsupervised Learning - Other Techniques

---

5. In-Class Assignment

• Objective:

  • Develop a classification model using a dataset of your choice.
  • Save the model as a pickle file.

• Steps:

  • Preprocess the data (handle missing values, encode categorical variables).
  • Train, evaluate and optimize the model.
  • Submit the pickle file of the trained model.

  Related notebook: In-Class Assignment

---

Using with GitHub Codespaces

You can run this repository directly in GitHub Codespaces without needing to set up anything locally.

Steps:

1. Click the Code button in the repository.
2. Select Open with Codespaces.
3. If you don’t see the option, click Create codespace on main.

Once the Codespace environment loads:

• All dependencies (e.g., Python packages) specified in requirements.txt will be automatically installed.
• You can open and run the Jupyter Notebooks directly in the Codespace.

---

Acknowledgements

Thanks to Leon Boschman for contributing his ideas, slides and feedback to this course material.