Machine learning is a branch of artificial intelligence (AI) that focuses on developing models and algorithms that allow computers to learn from data without explicit programming. In 1997, Mitchell defined Machine Learning as ‘the science of teaching computers to learn from data without being explicitly programmed’. Its primary objective is to train machines to simulate human-like thinking. Unlike traditional programming, where we give the computer explicit rules. Machine Learning uses algorithms that learn patterns from data and improve over time.
Branches of Machine Learning
Machine learning is majorly divided into the following branches:
-
Supervised Learning
-
Unsupervised Learning
- Reinforcement Learning
Supervised Learning
Supervised learning is a foundational approach in machine learning. In this method, the model is trained on labeled data, where both the input and the corresponding output are provided. Labeled data means that each input has a known output (outcome). For example, in student performance data, study hours and attendance are the inputs, while the output label specifies whether the student ‘passes’ or ‘fails’.
Unsupervised Learning
Unsupervised learning, on the other hand, works with data that has no labels. The model is not told what the correct answer is, it simply explores the data and tries to uncover patterns and relationships on its own. Two common ways this is done are clustering, where similar data points are grouped together, and dimensionality reduction, where the data is simplified by focusing only on its most important features. Popular algorithms include k-means for clustering and Principal Component Analysis (PCA) for dimensionality reduction. A practical example is in marketing, where companies use unsupervised learning to segment their customers. By analyzing purchase history, demographics, and behavior, the algorithm can naturally group customers with similar traits, even without predefined categories.
Reinforcement Learning
Reinforcement learning is a branch of machine learning that teaches an agent how to make decisions by interacting with its environment, much like how humans learn through trial and error. The agent takes actions, receives feedback in the form of rewards or penalties, and gradually learns which choices lead to better long-term outcomes. For example, when a baby touches a hot object, the pain acts as a penalty, and the baby quickly learns to avoid touching it again. Unlike supervised or unsupervised learning, reinforcement learning is especially useful when decisions build on one another where each step can influence what happens next. Everyday examples include learning to play a game, training a robot to walk, or managing resources efficiently, where success comes from balancing actions and rewards over time.
Machine Learning Pipeline
1. Preprocessing
- Identification of the problem
- Data collection
- Data cleaning
- Data preprocessing
- Feature extraction
- Feature engineering (if required)
- Exploratory Data Analysis
2. Model training
- Selection of machine learning model
- Evaluation of the model
3. Performance Matrix
- Confusion Matrix
- Precision, Recall and F1-Score
- AUC-ROC Curve
- Cross-validation
- Hyperparameter Tuning
