Specializing in Random Forest Implementations, Variable Selection, and OpenMP Parallel Processing
High-performance machine learning tools for researchers and developers
Fast Unified Random Forests for Survival, Regression, and Classification: This comprehensive package supports multiple analysis types including survival analysis, competing risks, multivariate regression, and class imbalanced classification. Features extreme random forests, advanced imputation methods, and innovative variable importance measures.
Super Greedy Trees for high-dimensional data analysis: This package features innovative splitting techniques including CART splits, hyperplane, ellipsoid and hyperboloid cuts. Uses coordinate descent for fast penalized lasso calculations and best split first (BSF) construction for optimal empirical risk reduction.
Model-Independent Variable Selection via Rule-Based Variable Priority. VarPro addresses bias in traditional permutation importance by using observed data comparisons within rule regions. This innovative approach provides robust variable selection for regression, classification, and survival analysis without artificial data manipulation.