def main_wc_gc_model():
config = ConfigParser.ConfigParser()
config.read('config/model_config.config')
app_log = setup_logging('wc_gc_logger',
config.get('logger', 'log_file_name'))
app_log.info('Scoring DUNS number: %d' % user_duns_number)
rate_lookup_table = read_rate_lookup(
config.get('data_files', 'rate_lookup'), user_is_uslh)
input_data['lookup_key'] = input_data['state'] + input_data['class_code']
avg_rates = merge_rate_lookup(input_data, rate_lookup_table)
entered_ratios = calc_entered_payroll_ratios(input_data)
inputs_valid, reason = check_inputs(input_data, entered_ratios)
if not inputs_valid:
return (numpy.NaN, numpy.NaN, numpy.NaN, reason)
payrolls = calc_payroll_ratio(input_data)
calc_allocate_clerical_payroll(payrolls, user_estimated_clerical_payroll)
calc_clerical_class_premium(payrolls, rate_lookup_table)
calc_standard_premium(payrolls, user_experience_mod)
standard_premium_to_allocate = calc_missing_standard_premium(
payrolls, avg_rates, user_experience_mod)
calc_allocated_standard_premium(payrolls, standard_premium_to_allocate)
calc_premium_discount(payrolls,
config.getfloat('constants', 'other_loadings'),
eval(config.get('data_files', 'ncci_tier_files')))
state_rate_data = merge_wcng_lr_rate_need(
payrolls, user_division, user_effective_date, user_is_uslh,
config.get('data_files', 'state_rate_need_lookup'),
config.get('data_files', 'wcng_lr'))
credit_scores = get_dnb_scores(
user_duns_number,
default_credit_score_pct=config.get('constants',
'default_duns_cs_pct'),
default_financial_score_pct=config.get('constants',
'default_duns_fs_pct'))
total_class_premium = input_data['class_premium'].sum()
predom_state = input_data.groupby(
by='state')['class_premium'].sum().idxmax(axis=1)
model_inputs = calc_normalized_claim_counts(
input_history, predom_state, eval(config.get('aqi', 'aqi_data')),
total_class_premium, config.get('data_files', 'cdf_file'))
model_inputs['credit_score_pct'] = credit_scores['credit_score_pct']
model_inputs['financial_score_pct'] = credit_scores['financial_score_pct']
model_inputs['payroll'] = user_total_projected_payroll
model_inputs['major_group'] = get_sic_major_group(user_sic_code)
predicted_lr = run_model(
model_inputs, config.get('data_files', 'model_coefficients_file'),
eval(config.get('model_rules', 'rules')))
state_rate_data['target_pricing_deviation_factor'] = (
((predicted_lr / state_rate_data['avg_wcng_loss_ratio']) *
state_rate_data['variable_rate_need']) +
state_rate_data['fix_rate_need'])
state_rate_data['estimated_premium'] = state_rate_data[
'target_pricing_deviation_factor'] * state_rate_data[
'manual_rate_pre_model']
output_midpoint = state_rate_data['estimated_premium'].sum()
lower_ratio, upper_ratio = calc_diamond_bound_ratios(
entered_ratios['clerical'], entered_ratios['non_clerical'],
config.get('data_files', 'bound_ratios'))
return (output_midpoint * lower_ratio, output_midpoint,
output_midpoint * upper_ratio, '')
def test_string_dnb(self):
with self.assertRaises(TypeError):
get_dnb_scores('123')
def test_bad_dnb(self):
scores = get_dnb_scores(123.456)
# Have to use assertIn since NaN == NaN gives False
self.assertIn(scores['credit_score_pct'], [NaN])
self.assertIN(scores['financial_score_pct'], [NaN])
def test_null_dnb(self):
with self.assertRaises(TypeError):
get_dnb_scores(None)
def test_vanilla(self):
scores = get_dnb_scores(608201141)
self.assertIn('credit_score_pct', scores)
self.assertIn('financial_score_pct', scores)
self.assertIsInstance(scores['credit_score_pct'], float)
self.assertIsInstance(scores['financial_score_pct'], float)