xgboost note

参数记录

param = {'bst:max_depth':3, 'bst:subsample':0.5, 'bst:min_child_weight':1,'bst:eta':0.3, 'silent':1,'objective':'binary:logistic'}
param['nthread'] = 2

1. 50 iter ：ａｕｃ：0.661716221418

param = {'bst:max_depth':3, 'bst:subsample':0.8, 'bst:min_child_weight':1,'bst:eta':0.01, 'silent':1,'objective':'binary:logistic'}
    param['nthread'] = 2
    # banlance
    #param['scale_pos_weight'] = 1
    # auc
    param['eval_metric'] = 'auc'

1. 
   0.661716221418

    # setting patameters
    param = {'bst:subsample':0.8, 'bst:min_child_weight':1, 'silent':1,'objective':'binary:logistic'}
    param['nthread'] = 2
    # banlance
    #param['scale_pos_weight'] = 1
    # auc
    param['eval_metric'] = 'auc'
    # important feature
    param['bst:max_depth'] = 6
    param['bst:min_child_weight'] = 1
    param['bst:eta'] = 0.1

    # cross validation
    #cross_validation(DATA_PATH+"processed",param)
    # num_round

1. 
   [49] eval-auc:0.661716 train-auc:0.670260

    # setting patameters
    param = {'bst:subsample':0.8, 'bst:min_child_weight':1, 'silent':1,'objective':'binary:logistic'}
    param['nthread'] = 2
    # banlance
    #param['scale_pos_weight'] = 1
    # auc
    param['eval_metric'] = 'auc'
    # important feature
    param['bst:max_depth'] = 10
    param['bst:min_child_weight'] = 1
    param['bst:eta'] = 0.3

    # cross validation
    #cross_validation(DATA_PATH+"processed",param)
    # num_round
    num_round = 50

1. 增加label

[Dimension]
idfa_names=id,city,street,system_info,version,dpi,tag1,tag2,tag3,tag4,tag5,tag6,tag7
imei_names=id,androidid,mac,city,street,system_info,version,dpi,tag1,tag2,tag3,tag4,tag5,tag6,tag7

[Parameter]
extention=1
processed_name=processed_extention.data
model_name=0001.model
nthread=3
max_depth=20
min_child_weight=1
eta=0.3
num_round=100
~
~

[94]    eval-auc:0.660827       train-auc:0.679268
[95]    eval-auc:0.660816       train-auc:0.679359
[96]    eval-auc:0.660762       train-auc:0.679443
[97]    eval-auc:0.660748       train-auc:0.679526
[98]    eval-auc:0.660743       train-auc:0.679559
[99]    eval-auc:0.660754       train-auc:0.679658
     feature  code
208  2011117   166
223  2011301    97
230  2011308    92
19   1010301    91
203  2011112    83
222    20113    76
99     10305    74
156    20107    73
152  2010601    69
244  2011411    68

5。linear 无正则化

[95]    eval-auc:0.623535       train-auc:0.623284
[96]    eval-auc:0.623539       train-auc:0.623282
[97]    eval-auc:0.623535       train-auc:0.623281
[98]    eval-auc:0.623536       train-auc:0.623280
[99]    eval-auc:0.623534       train-auc:0.623279