Clustering Analysis Using Survival Tree and Forest Algorithms

An outcome-guided algorithm is developed to identify clusters of samples with similar characteristics and survival rate. The algorithm first builds a random forest and then defines distances between samples based on the fitted random forest. Given the distances, we can apply hierarchical clustering algorithms to define clusters. Details about this method is described in < https://github.com/luyouepiusf/SurvivalClusteringTree>.

Implements Under/Oversampling for Probability Estimation

Implements under/oversampling for probability estimation. To be used with machine learning methods such as AdaBoost, random forests, etc.

Multivariate Outlier Detection and Replacement

Provides a random forest based implementation of the method described in Chapter 7.1.2 (Regression model based anomaly detection) of Chandola et al. (2009) . It works as follows: Each numeric variable is regressed onto all other variables by a random forest. If the scaled absolute difference between observed value and out-of-bag prediction of the corresponding random forest is suspiciously large, then a value is considered an outlier. The package offers different options to replace such outliers, e.g. by realistic values found via predictive mean matching. Once the method is trained on a reference data, it can be applied to new data.

A General Iterative Clustering Algorithm

An iterative algorithm that improves the proximity matrix (PM) from a random forest (RF) and the resulting clusters as measured by the silhouette score.

Interpret Tree Ensembles

For tree ensembles such as random forests, regularized random forests and gradient boosted trees, this package provides functions for: extracting, measuring and pruning rules; selecting a compact rule set; summarizing rules into a learner; calculating frequent variable interactions; formatting rules in latex code. Reference: Interpreting tree ensembles with inTrees (Houtao Deng, 2019, ).

Out of Bag Learning Curve

Provides functions to calculate the out-of-bag learning curve for random forests for any measure that is available in the 'mlr' package. Supported random forest packages are 'randomForest' and 'ranger' and trained models of these packages with the train function of 'mlr'. The main function is OOBCurve() that calculates the out-of-bag curve depending on the number of trees. With the OOBCurvePars() function out-of-bag curves can also be calculated for 'mtry', 'sample.fraction' and 'min.node.size' for the 'ranger' package.

Impute Missing Values in a Predictive Context

Compute missing values on a training data set and impute them on a new data set. Current available options are median/mode and random forest.

Stable and Interpretable RUle Set

A regression and classification algorithm based on random forests, which takes the form of a short list of rules. SIRUS combines the simplicity of decision trees with a predictivity close to random forests. The core aggregation principle of random forests is kept, but instead of aggregating predictions, SIRUS aggregates the forest structure: the most frequent nodes of the forest are selected to form a stable rule ensemble model. The algorithm is fully described in the following articles: Benard C., Biau G., da Veiga S., Scornet E. (2021), Electron. J. Statist., 15:427-505 for classification, and Benard C., Biau G., da Veiga S., Scornet E. (2021), AISTATS, PMLR 130:937-945 < http://proceedings.mlr.press/v130/benard21a>, for regression. This R package is a fork from the project ranger (< https://github.com/imbs-hl/ranger>).

Transformation Trees and Forests

Recursive partytioning of transformation models with corresponding random forest for conditional transformation models as described in 'Transformation Forests' (Hothorn and Zeileis, 2021, ) and 'Top-Down Transformation Choice' (Hothorn, 2018, ).

Implements the Forest-R.K. Algorithm for Classification Problems

Provides functions that calculates common types of splitting criteria used in random forests for classification problems, as well as functions that make predictions based on a single tree or a Forest-R.K. model; the package also provides functions to generate importance plot for a Forest-R.K. model, as well as the 2D multidimensional-scaling plot of data points that are colour coded by their predicted class types by the Forest-R.K. model. This package is based on: Bernard, S., Heutte, L., Adam, S., (2008, ISBN:978-3-540-85983-3) "Forest-R.K.: A New Random Forest Induction Method", Fourth International Conference on Intelligent Computing, September 2008, Shanghai, China, pp.430-437.