xgboost_default.py

#!/usr/bin/env python3
 
 
 
import basf2_mva
import basf2_mva_util
import time
 
if __name__ == "__main__":
    from basf2 import conditions, find_file
    # NOTE: do not use testing payloads in production! Any results obtained like this WILL NOT BE PUBLISHED
    conditions.testing_payloads = [
        'localdb/database.txt'
    ]
 
    variables = ['M', 'p', 'pt', 'pz',
                 'daughter(0, p)', 'daughter(0, pz)', 'daughter(0, pt)',
                 'daughter(1, p)', 'daughter(1, pz)', 'daughter(1, pt)',
                 'daughter(2, p)', 'daughter(2, pz)', 'daughter(2, pt)',
                 'chiProb', 'dr', 'dz',
                 'daughter(0, dr)', 'daughter(1, dr)',
                 'daughter(0, dz)', 'daughter(1, dz)',
                 'daughter(0, chiProb)', 'daughter(1, chiProb)', 'daughter(2, chiProb)',
                 'daughter(0, kaonID)', 'daughter(0, pionID)',
                 'daughterInvM(0, 1)', 'daughterInvM(0, 2)', 'daughterInvM(1, 2)']
 
    train_file = find_file("mva/train_D0toKpipi.root", "examples")
    test_file = find_file("mva/test_D0toKpipi.root", "examples")
 
    training_data = basf2_mva.vector(train_file)
    testing_data = basf2_mva.vector(test_file)
 
    general_options = basf2_mva.GeneralOptions()
    general_options.m_datafiles = training_data
    general_options.m_treename = "tree"
    general_options.m_identifier = "XGBoost"
    general_options.m_variables = basf2_mva.vector(*variables)
    general_options.m_target_variable = "isSignal"
 
    specific_options = basf2_mva.PythonOptions()
    specific_options.m_steering_file = 'mva/examples/python/xgboost_default.py'
    specific_options.m_framework = "xgboost"
    param = ('{"max_depth": 3, "eta": 0.1, "silent": 1, "objective": "binary:logistic",'
             '"subsample": 0.5, "nthread": 1, "nTrees": 100}')
    specific_options.m_config = param
 
    training_start = time.time()
    basf2_mva.teacher(general_options, specific_options)
    training_stop = time.time()
    training_time = training_stop - training_start
    method = basf2_mva_util.Method(general_options.m_identifier)
    inference_start = time.time()
    p, t = method.apply_expert(testing_data, general_options.m_treename)
    inference_stop = time.time()
    inference_time = inference_stop - inference_start
    auc = basf2_mva_util.calculate_auc_efficiency_vs_background_retention(p, t)
    print("XGBoost", training_time, inference_time, auc)