Advanced Methods in Data Science

Machine Learning

What is it? (discuss)

What is Machine Learning?

• subfield of computer science + statistics
• building algorithms/models based on data (i.e. collection of examples of some phenomenon)

Types of Machine Learning

• supervised (labeled data)
• semi-supervised (both labeled and unlabeled data)
• unsupervised (unlabeled data)
• reinforcement (reward system, with the goal of learning a policy)

For all of these, data is required – one important step is feature engineering (how to transform data into feature – more on this later)

Depending on the target type, classification or regression

Supervised Learning

• Labeled data – input (features), output (target, response)

Can you think of examples of supervised learning?

Supervised Learning

• Example: spam vs. not spam

What features can we use?

Supervised Learning Algorithms

• Support Vector Machines (SVM)
• Decision Tree/Random Forest
• Logistic Regression
• Linear Regression
• Naive Bayes
• K-Nearest Neighbors

How to decide?

Unsupervised Learning

• Clustering (K-Means)
• Principal Component Analysis (Dimensionality Reduction)

Reinforcement Learning

• Recommendation Systems (users give feed on whether a recommendation was good or not)
• Automated Robots
• Autonomous Driving
• Natural Language Processing – text prediction, translation