my xgboost configuration for kaggle-otto competition

otto.conf

# local-CV: 0.436

# number of threads
nthread=8
# fixed random seed
seed=12
# whether create binary buffer for text input, this normally will speedup loading (NO)
use_buffer = 0

# The path of training data
model_out = "test_seed12.model"
data = "./data/train.libsvm"
# The path of test data
test:data = "./data/test.libsvm"

# 0 means do not save any model except the final round model
save_period = 0

# Boosting settings

# Tree booster
booster = gbtree
# the initial prediction score of all instances, global bias. (0.11521 = mean CTR)
# subsample ratio of the training instance
subsample = 1.0
colsample_bytree = 0.3
num_parallel_tree = 4 # boosted random forest

# step size / shrinkage
eta = 0.025
# minimum loss reduction required to make a further partition
gamma = 0.5
min_child_weight = 4

# maximum depth of a tree
max_depth = 10
# the number of round to do boosting
num_round = 3000

# objective function / evaluation metric
objective = multi:softprob
num_class = 9
eval_metric = mlogloss

# evaluate on training data as well each round
eval_train = 1