1. Supervised Learning:
- Description: Model is trained with labeled data (inputs with known outputs).
- Types of Supervised Problems:
- Classification: Target values are categorical (discrete).
- Binary Classification: Example: Is it a fish or not a fish?
- Multiclass Classification: Example: Categorize incoming customer feedback/reviews into one of several topics, such as Product Quality, Customer Service, Shipping Issues, and Pricing
Example: Document could be classified as being about religion, politics, or finance etc.
- Regression: Target values are continuous (mathematical).
- Example: Predicting house prices based on features like square footage, number of bedrooms, etc.
2. Unsupervised Learning:
- Description: Model is trained with data that has no labeled target values. The model discovers patterns within the data.
- Types of Unsupervised Problems:
- Clustering: Grouping similar data points.
- Example: Group customers by purchase history.
- Anomaly Detection: Identifying outliers or rare events.
- Example: Detecting fraud or sensor malfunctions.
Supervised Learning Details
Classification:
- Binary Classification: Two categories.
- Example: Predict whether a person has a disease or not.
- Multiclass Classification: More than two categories.
- Example: Predict the topic of a tax document (religion, politics, etc.).
Regression:
- Linear Regression: Predict a continuous value with a linear relationship.
- Example: Predict a person’s height based on weight.
- Multiple Linear Regression: Predict using multiple factors.
- Example: Predict house prices using features like square footage, number of bedrooms, etc.
- Logistic Regression: Predict the probability of an event happening.
- Example: Predict if a person will get heart disease based on BMI, smoking status, etc.
Unsupervised Learning Details
Clustering:
- Groups similar data points together.
- Example: Grouping customers by purchase history.
Anomaly Detection:
- Detects rare events or outliers.
- Example: Identifying fraudulent transactions.
