| name | r-ml |
| description | R machine learning packages. Use for classification, regression, clustering, deep learning, gradient boosting (xgboost, lightgbm), random forests, neural networks, and time series forecasting. |
R Machine Learning Skill
Sub-skills
| Sub-skill | Description |
|---|
| r-ml-frameworks | tidymodels, caret, mlr3, h2o |
| r-ml-boosting | xgboost, lightgbm, gbm |
| r-ml-trees | randomForest, ranger, rpart |
| r-ml-regularization | glmnet, lasso, elastic-net |
| r-ml-deeplearning | torch, keras, neural networks |
| r-ml-timeseries | prophet, fable, forecast |
| r-ml-survival | survival, survminer |
| r-ml-anomaly | AnomalyDetection, anomalize |
Machine learning and predictive modeling in R.
ML Frameworks
| Package | Description |
|---|
| caret ★ | Classification and Regression Training |
| mlr3 ★ | Next-gen extensible ML framework |
| tidymodels ★ | Tidyverse-friendly modeling |
| h2o | Deep learning, RF, GBM, GLM |
Gradient Boosting
| Package | Description |
|---|
| xgboost ★ | eXtreme Gradient Boosting |
| lightgbm ★ | Light Gradient Boosting Machine |
| gbm | Generalized Boosted Regression |
| bst | Gradient Boosting |
| mboost | Model-Based Boosting |
| CoxBoost | Cox models boosting |
| GAMBoost | GAM boosting |
| gamboostLSS | GAMLSS boosting |
| GMMBoost | Mixed models boosting |
Tree-Based Methods
| Package | Description |
|---|
| randomForest | Breiman's random forests |
| ranger ★ | Fast random forests |
| randomForestSRC | RF for survival/regression/classification |
| rpart | Recursive partitioning trees |
| party | Recursive partitioning lab |
| partykit | Partitioning toolkit |
| C50 | C5.0 Decision Trees |
| Cubist | Rule-based regression |
| evtree | Evolutionary trees |
| tree | Classification/regression trees |
| bigrf | Big Random Forests |
Regularization & Linear Models
| Package | Description |
|---|
| glmnet ★ | Lasso and elastic-net GLMs |
| lars | Least Angle Regression, Lasso |
| elasticnet | Elastic-Net, Sparse PCA |
| penalized | L1/L2 penalized estimation |
| ncvreg | SCAD/MCP regularization |
| grplasso | Group Lasso |
| grpreg | Grouped covariates regularization |
| L0Learn | Best subset selection |
Neural Networks & Deep Learning
| Package | Description |
|---|
| torch ★ | PyTorch-like tensors/NNs |
| MXNet | Flexible GPU deep learning |
| nnet | Feed-forward NNs |
| RSNNS | Stuttgart NN Simulator |
| keras | Keras interface |
Mixed Effects Models
| Package | Description |
|---|
| lme4 ★ | Mixed-effects models |
| nlme | Mixed-effects with custom covariance |
| glmmTMB | Generalized mixed-effects |
Time Series & Forecasting
| Package | Description |
|---|
| prophet ★ | Facebook's forecasting tool |
| fable | Tidy forecasting |
| forecast | Time series forecasting |
Anomaly Detection
| Package | Description |
|---|
| AnomalyDetection | Twitter's anomaly detection |
| anomalize | Tidy anomaly detection |
| BreakoutDetection | Twitter's breakout detection |
| CausalImpact | Google's causal inference |
SVM & Kernel Methods
| Package | Description |
|---|
| kernlab | Kernel-based ML lab |
| e1071 | SVM, Naive Bayes, etc. |
| LiblineaR | Linear predictive models |
| svmpath | SVM path algorithm |
| penalizedSVM | Feature selection SVM |
Clustering & Dimensionality
| Package | Description |
|---|
| kohonen | Self-Organizing Maps |
| Rsomoclu | Parallel SOM |
| hda | Heteroscedastic Discriminant Analysis |
| klaR | Classification and visualization |
Feature Selection
| Package | Description |
|---|
| Boruta | All-relevant feature selection |
| FSelector | Subset-search/ranking selection |
| varSelRF | Variable selection with RF |
Survival Analysis
| Package | Description |
|---|
| survival ★ | Survival analysis |
| survminer | Survival visualization |
| ahaz | Additive hazards regression |
Other
| Package | Description |
|---|
| arules | Association rules mining |
| rattle | GUI for data mining |
| rminer | Simplified NN/SVM usage |
| ROCR | Classifier performance visualization |
| SuperLearner | Ensemble learning |
| RWeka | Weka interface |
Quick Examples
library(tidymodels)
split <- initial_split(df, prop = 0.8)
train <- training(split)
test <- testing(split)
model <- rand_forest(trees = 100) %>%
set_engine("ranger") %>%
set_mode("classification")
recipe <- recipe(target ~ ., data = train) %>%
step_normalize(all_numeric())
workflow <- workflow() %>%
add_model(model) %>%
add_recipe(recipe)
fit <- workflow %>% fit(data = train)
predict(fit, test)
library(xgboost)
dtrain <- xgb.DMatrix(data = as.matrix(train_x), label = train_y)
model <- xgb.train(
params = list(objective = "binary:logistic", max_depth = 6),
data = dtrain, nrounds = 100
)
library(caret)
ctrl <- trainControl(method = "cv", number = 5)
model <- train(target ~ ., data = train, method = "rf", trControl = ctrl)
Resources
