import xgboost as xgb
from sklearn.grid_search import ParameterGrid
from sklearn.grid_search import ParameterSampler
import pandas as pd
import matplotlib.pyplot as plt
import logging
class XGBoostBase(object):
"""Base class for XGBoost learning model.
It implements some common functionality, including model training/validation, cross valiation.
Derived class need to override set_xgb_parameters method to specify xgboost parameters.
To use this class properly, it must be guarantted that its derived class also inherit preparedata class which
has interface get_train_validationset and getFeaturesLabel
"""
def __init__(self):
self.do_cross_val = True
self.best_score_colname_in_cv = 'test-mape-mean'
return
def run(self):
# self.get_model_input()
if self.do_cross_val is None:
return self.run_grid_search()
if self.do_cross_val:
return self.run_croos_validation()
return self.run_train_validation()
def run_croos_validation(self):
features,labels,cv_folds = self.getFeaturesLabel()
dtrain_cv = xgb.DMatrix(features, label= labels,feature_names=features.columns)
self.set_xgb_parameters()
# specify validations set to watch performance
model = xgb.cv(self.xgb_params, dtrain_cv, folds=cv_folds, **self.xgb_learning_params)
best_scroe = model[self.best_score_colname_in_cv].max()
return best_scroe
def set_xgb_parameters(self):
self.xgb_params = {'silent':1}
self.xgb_learning_params = {}
return
def run_train_validation(self):
x_train, y_train,x_validation,y_validation = self.get_train_validationset()
dtrain = xgb.DMatrix(x_train, label= y_train,feature_names=x_train.columns)
dvalidation = xgb.DMatrix(x_validation, label= y_validation,feature_names=x_validation.columns)
self.set_xgb_parameters()
evals=[(dtrain,'train'),(dvalidation,'eval')]
model = xgb.train(self.xgb_params, dtrain, evals=evals, **self.xgb_learning_params)
xgb.plot_importance(model)
plt.show()
print "features used:\n {}".format(self.get_used_features())
return
class XGBoostGridSearch(object):
"""Base class for XGBoost learning model.
It implements parameter grid search functionality
To use this class properly, it must be guarantted that its derived class also inherit preparedata class which
has interface get_train_validationset and getFeaturesLabel
"""
def __init__(self):
self.ramdonized_search_enable = False
self.randomized_search_n_iter = 10
self.ramdonized_search_random_state = None
self.grid_search_display_result = True
self.__grid_search_result = []
return
def get_paramgrid_1(self):
pass
def get_paramgrid_2(self, param_grid):
return param_grid
def get_learning_params(self):
pass
def __get_param_iterable(self, param_grid):
if self.ramdonized_search_enable:
parameter_iterable = ParameterSampler(param_grid,
self.randomized_search_n_iter,
random_state=self.ramdonized_search_random_state)
else:
parameter_iterable = ParameterGrid(param_grid)
return parameter_iterable
def __get_param_grid(self):
param_grid = self.get_paramgrid_1()
param_grid = self.get_paramgrid_2(param_grid)
return param_grid
def run_grid_search(self):
"""
This method is called by derived class to start grid search process
"""
features,labels,cv_folds = self.getFeaturesLabel()
dtrain_cv = xgb.DMatrix(features, label= labels,feature_names=features.columns)
parameter_iterable = self.__get_param_iterable(self.__get_param_grid())
kwargs = self.get_learning_params()
for param in parameter_iterable:
logging.info("used parameters: {}".format(param))
bst = xgb.cv(param, dtrain_cv, folds=cv_folds,**kwargs)
self.__add_to_resultset(param, bst)
self.__disp_result()
return
def __add_to_resultset(self, param, bst):
max_id = bst[self.best_score_colname_in_cv].idxmax()
self.__grid_search_result.append((param, bst.iloc[max_id][self.best_score_colname_in_cv], bst.iloc[max_id].tolist()))
logging.info("CV score: {}".format(bst.iloc[max_id]))
return
def __disp_result(self):
if not self.grid_search_display_result:
return
df = pd.DataFrame(self.__grid_search_result, columns= ['param', 'result', 'otherinfo'])
logging.info( '\nall para search results:')
logging.info("{}".format( df.values))
best_score_id = df['result'].idxmax()
logging.info( '\nbest parameters:')
# logging.info("{}".format(df.iloc[best_score_id]['param']))
# logging.info("{}".format( df.iloc[best_score_id]['result']))
logging.info("{}".format( df.iloc[best_score_id].values))
df.to_csv('temp/__grid_search_result.csv')
return
