I have trained an XGBregressor model with following parameters:
{‘objective’: ‘reg:gamma’,
‘base_score’: 0.5,
‘booster’: ‘gbtree’,
‘colsample_bylevel’: 0.7,
‘colsample_bynode’: 1,
‘colsample_bytree’: 0.7,
‘eval_metric’: None,
‘gamma’: 0,
‘gpu_id’: -1,
‘grow_policy’: ‘depthwise’,
‘interaction_constraints’: ‘’,
‘learning_rate’: 0.02,
‘max_bin’: 256,
‘max_cat_to_onehot’: 4,
‘max_delta_step’: 0,
‘max_depth’: 17,
‘max_leaves’: 0,
‘min_child_weight’: 2,
‘monotone_constraints’: ‘(1,-1,0,-1,-1,-1,0,-1,0,0)’,
‘n_jobs’: 30,
‘num_parallel_tree’: 1,
‘predictor’: ‘auto’,
‘random_state’: 13,
‘reg_alpha’: 1.5,
‘reg_lambda’: 1,
‘sampling_method’: ‘uniform’,
‘scale_pos_weight’: None,
‘subsample’: 0.9,
‘tree_method’: ‘exact’,
‘validate_parameters’: 1,
‘verbosity’: None}
But monotonic constraint does not work as expected, I mean for a fixed feature set by changing a monotone feature the result was not monotonic.
Would you please help me how can I solve this issue?