Data Science with R: Machine Learning

This 35-hour Machine Learning with R course introduces both the theoretical foundation of machine learning algorithms as well as their practical applications in R. It will introduce you to data mining, performance measures and dimension reduction, regression models, both linear and generalized, KNN and Naïve Bayes models, tree models, and SVMs as well as the Association Rule for analysis. After successfully completing of this course, you will be able to break down the mathematics behind major machine learning algorithms, explain the principles of machine learning algorithms, and implement these methods to solve real-world problems.

Prerequisites:

• Knowledge of R programming
• Able to munge, analyze, and visualize data in R

Syllabus:
Unit 1: Foundations of Statistics and Simple Linear Regression

• Understand your data
• Statistical inference
• Introduction to machine learning
• Simple linear regression
• Diagnostics and transformations
• The coefficient of determination

Unit 2: Multiple Linear Regression and Generalized Linear Model

• Multiple linear regression
• Assumptions and diagnostics
• Extending model flexibility
• Generalized linear models
• Logistic regression
• Maximum likelihood estimation
• Model interpretation
• Assessing model fit

Unit 3: kNN and Naive Bayes, the Curse of Dimensionality

• The K-Nearest Neighbors Algorithm
• The choice of K and distance measure
• Conditional probability: Bayes’ Theorem
• The Naive Bayes’ Algorithm
• The Laplace estimator
• Dimension reduction
• The PCA procedure
• Ridge and Lasso regression
• Cross-validation

Unit 4: Tree Models and SVMs

• Decision trees
• Bagging
• Random forests
• Boosting
• Variable Importance
• Hyperplanes and maximal margin classifier
• Sort margin and support vector classifier
• Kernels and support vector machines

Unit 5: Cluster Analysis and Neural Networks

• Cluster analysis
• K-means clustering
• Hierarchical clustering
• Neural networks and perceptrons
• Sigmoid neurons
• Network topology and hidden features
• Back propagation learning with gradient descent