Regression Tree implimenting CART algorithm

Classification Tree implimenting CART algorithm

Base Class for implimenting Linear Regression

Base Tree for classification/regression. Written for single variable/value binary split critereon. Many methods needs to be rewritten i

Bernoulli Naive Bayes Classifier

Discriminant Adaptive Nearest Neighbors (DANN). DANN adaptively elongates neighborhoods along boundry regions. Useful for high dimensiona

Ada Boost classifier. This implimentation produces a series of decisions stumps (decision trees with two terminal nodes).

Class for implimenting Regularized Discriminent Analysis LDA is performed when alpha=0 QDA is performed when alpha=1

Common error measures for model predictions when true answer is known

Gaussian Mixture classification

Gaussian Naive Bayes Classifier

Gradient boosting regression. A series of small trees (weak classifiers) combined to produce a hypothesis. Each subsequent classifier fi

K means clustering

K mediods clustering. Similar to K means, however, datapoints (mediods/exemplars) are chosen as cluster centers. TODO: merge in with

K nearest neighbors classification and regression

Kernel methods for classification and estimation

Learning Vector Quantization: Prototypes are attracted to training points of correct class, and repeled from training points in incorrect

Class for implimenting Linear Regression

Logistic Regression classifier with gradient descent implementation

Naive Bayes Classifier Given class label, assumes features are independent

Exact principal component analysis (PCA). Transforms given data set into orthonormal basis, maximizing variance. TODO return variance ex

Perceptron classifier of binary data. If data is not linearly seperable data, the pocket algorithm is highly recommended

PRIM: Patient Rule Induction Method Decision at node peels of 10% of data which maximizes response mean More "patient" than CART algorit

Random Forests are an ensemble method which averages a set of de-correlated trees. In this implimentation, a bootstrapped sample is used