XGBoost Classifier Lab

Upload your own CSV (or pick a sample dataset), tune the gradient boosting hyper-parameters — or keep the defaults — and train a regularised gradient boosted tree ensemble (XGBoost-style: second-order gradients, shrinkage, subsampling, L1/L2 regularisation) directly in your browser.

1 · Data

Sample dataset

…or upload a CSV (last column = label)

Paste CSV instead

2 · XGBoost parameters

Leave everything untouched to train with the standard defaults.

n_estimators (boosting rounds)

Number of trees. Default 150.

learning_rate (eta)

Shrinkage per tree. Default 0.3.

max_depth

Tree depth. Default 4.

min_child_weight

Min sum of hessians in a leaf. Default 1.

gamma (min_split_loss)

Min gain to split. Default 0.

lambda (L2)

L2 leaf regularisation. Default 1.

alpha (L1)

L1 leaf regularisation. Default 0.

subsample

Row sampling per tree. Default 1.

colsample_bytree

Feature sampling per tree. Default 1.

test size

Stratified hold-out fraction. Default 0.25.

random_state

Seed for reproducibility.

early_stopping_rounds (0 = off)

Stop when test loss stalls.

XGBoost Classifier Lab

1 · Data

Class distribution

Preview (first 5 rows)

2 · XGBoost parameters

3 · Results

Confusion matrix (test set)

Per-class report (test set)

Feature importance (total gain)

Training curve (log-loss)