Machine Learning and Statistical Modeling with R Examples

Explore how to assess model accuracy using cross-validation, training vs test data, and common validation methods like hold-out, k-fold, leave-one-out, and bootstrapping.

Learn to perform cross validation in R with a reproducible workflow, from seed setting and random sampling to linear modeling and comparing leave-one-out and k-fold validation using bootstrapping tools.

Learn cross-validation with the faithful dataset, analyzing waiting and eruptions to reveal a linear relationship, and compare simple cross-validation with five-fold cross-validation using a 50-50 split.

Explore the exercise database at Armani's tutorials dot com, with blocks of ten questions. Select topics like functions, data frame handling, and graphs, attempt A–C questions, and view solutions after.

Examine classification methods in machine learning, including logistic regression, linear discriminant analysis, k-nearest neighbor, and support vector machines, for quantitative and qualitative data with probabilistic outputs.

Use k-nearest neighbors to classify cylinder count from mpg and weight in the Motor Trend car road tests dataset, with train and test data and vote proportions.

Learn to perform linear discriminant analysis in R with the LDA function using a weight and mpg formula. Interpret priors, means, and coefficients; predict on test data to assess accuracy.

Explore logistic regression for classification in R, modeling a binary transmission type with a glm binomial, using weight and mpg, and evaluate training predictions and 0.5 thresholds.

Classify iris species using petal width and length with linear discriminant analysis and k-nearest neighbors (k=3), train on iris data, and validate predictions on a test set.

Explore decision trees for classification and regression with both qualitative and quantitative predictors and outcomes. Learn how pruning, bagging, random forests, and boosting improve accuracy on complex data.

Explore building and visualizing decision trees in R with the tree package, using a simple formula for weight, mpg, and transmission type to create regression, pseudo regression, and classification trees.

Train a classification tree on mpg and am, predict on the test set, and report 78.5% accuracy with 13% misclassification, noting pruning or boosting options.

Explore how bagging and random forests use bootstrapping and limited predictor variables to reduce variance, build multiple trees, and compare their accuracy on the thiamin dataset using R's randomForest package.

Build a classification tree on the diamante dataset using price and index to predict color. Use 500 rows with 250/250 split and note ~17% accuracy; consider boosting or random forests.

Cluster analysis is an unsupervised learning method that identifies patterns and forms clusters within data, with k-means needing predefined k and hierarchical clustering using dendrograms and Euclidean distance.

Explore k-means clustering in r using the base package to partition the reverse dataset into three clusters, examining centers, cluster means, and plotting the clustered results.

Explore hierarchical clustering in R by calculating euclidean distances, applying complete, single, and average linkage, and interpreting the resulting dendrogram to assess similarity among observations.

Apply k-means clustering to iris data using sepal length, sepal width, and petal length, compare k = 3, 5, and 8, and visualize 3D cluster memberships.