def __init__(self, start_date, end_date, fake_today):
engine, config = setup_environment.get_database()
self.con = engine.raw_connection()
self.con.cursor().execute("SET SCHEMA '{}'".format(config['schema']))
self.start_date = start_date
self.end_date = end_date
self.fake_today = fake_today
self.tables = config # Dict of tables
self.schema = config['schema']
import numpy as np
import pdb
import pandas as pd
import yaml
import logging
import sys
import datetime
from eis import setup_environment
from eis.features import class_map
log = logging.getLogger(__name__)
engine, config = setup_environment.get_database()
con = engine.raw_connection()
con.cursor().execute("SET SCHEMA '{}'".format(config["schema"]))
def get_baseline(ids, start_date, end_date):
"""
Gets EIS baseline - whether or not an officer is flagged
by the EIS at any point in the labelling window.
Returns dataframe with ids and boolean value corresponding to if
an officer was flagged or not.
"""
flagged_officers = (
"select distinct newid from {} " "WHERE datecreated >= '{}'::date " "AND datecreated <='{}'::date"
).format(config["eis_table"], start_date, end_date)
labels = pd.read_sql(flagged_officers, con=con)
Code to take top performing recent models and
put them in the evaluation webapp for further
examination.
Examples:
--------
python prepare.py '2016-08-03' 'auc'
python prepare.py '2016-08-03' 'recall@' -p '0.01'
python prepare.py '2016-08-03' 'precision@' -p '10.0' -n 10
python prepare.py '2016-08-03' 'precision@' -p '10.0' -n 10 -d 'example_directory/'
"""
engine = setup_environment.get_database()
try:
con = engine.raw_connection()
con.cursor().execute("SET SCHEMA '{}'".format('models'))
except:
pass
def get_metric_best_models(timestamp, metric, parameter=None, number=25):
"""
--------------------------------------------------------
Get the EVALUATION METRIC VALUE of the best models
by the specified timestamp and given metric
--------------------------------------------------------
ARGUMENTS:
timestamp: models run on or after given timestamp
