XGBoost.jl
eXtreme Gradient Boosting Package in Julia
Abstract
This package is a Julia interface of XGBoost, which is short for eXtreme gradient Gradient Boosting. It is an efficient and scalable implementation of gradient boosting framework.The package includes efficient linear model solver and tree learning algorithms. The library is parallelized using OpenMP, and it can be more than 10 times faster some of than existing gradient boosting packages. It supports various objective functions, including regression, classification and ranking. The package is also made to be extensible, so that users are also allowed to define their own objectives easily.
Features
- Sparse feature format, it allows easy handling of missing values, and improve computation efficiency.
- Advanced features, such as customized loss function, cross validation, see demo folder for walkthrough examples.
Installation
Pkg.add("XGBoost")
or
Pkg.clone("https://github.com/antinucleon/XGBoost.jl.git")
Pkg.build("XGBoost")
The XGBoost
package also depends on the BinDeps
Minimal examples
To show how XGBoost works, here is an example of dataset Mushroom
- Prepare Data
XGBoost support Julia Array
, SparseMatrixCSC
, libSVM format text and XGBoost binary file as input. Here is an example of Mushroom classification. This example will use the function readlibsvm
in basic_walkthrough.jl. This function load libsvm format text into Julia dense matrix.
using XGBoost
train_X, train_Y = readlibsvm("data/agaricus.txt.train", (6513, 126))
test_X, test_Y = readlibsvm("data/agaricus.txt.test", (1611, 126))
- Fit Model
num_round = 2
bst = xgboost(train_X, num_round, label=train_Y, eta=1, max_depth=2)
Predict
pred = predict(bst, test_X)
print("test-error=", sum((pred .> 0.5) .!= test_Y) / float(size(pred)[1]), "\n")
Cross-Validation
nfold=5
param = ["max_depth"=>2, "eta"=>1, "objective"=>"binary:logistic"]
metrics = ["auc"]
nfold_cv(train_X, num_round, nfold, label=train_Y, param=param, metrics=metrics)
Feature Walkthrough
Check demo
- Basic walkthrough of features
- Cutomize loss function, and evaluation metric
- Boosting from existing prediction
- Predicting using first n trees
- Generalized Linear Model
- Cross validation
Model Parameter Setting
Check XGBoost Wiki