# DEALINGS IN THE SOFTWARE.
import numpy as np
import pandas as pd
import getml.models.aggregations as aggregations
import getml.datasets as datasets
import getml.engine as engine
import getml.models.loss_functions as loss_functions
import getml.models as models
import getml.data as data
import getml.predictors as predictors
# ----------------
engine.set_project("examples")
# ----------------
# Generate artificial dataset
# The problem we create looks like this:
#
# SELECT COUNT( * )
# FROM POPULATION t1
# LEFT JOIN PERIPHERAL t2
# ON t1.join_key = t2.join_key
# WHERE (
# ( t1.time_stamp - t2.time_stamp <= 0.5 )
# ) AND t2.time_stamp <= t1.time_stamp
# GROUP BY t1.join_key,
# t1.time_stamp;
def test_relboost_same_units():
"""Check if the same results will be obtained regardless of whether
the units are assigned to the DataFrame, to the Columns, or to the
RelboostModel.
"""
# ----------------------------------------------------------------
engine.set_project("examples")
seed = 33231
units = {"column_01": "column_01"}
# ----------------------------------------------------------------
# Generate artificial dataset
# The problem we create looks like this:
#
# SELECT COUNT( * )
# FROM POPULATION t1
# LEFT JOIN PERIPHERAL t2
# ON t1.join_key = t2.join_key
# WHERE (
# ( t1.time_stamp - t2.time_stamp <= 0.5 )
# ) AND t2.time_stamp <= t1.time_stamp
# GROUP BY t1.join_key,
# t1.time_stamp;
# ----------------------------------------------------------------
# Assign the units to the columns
population_table_columns, peripheral_table_columns = datasets.make_same_units_numerical(
random_state=seed)
population_table_columns.set_unit("column_01", "column_01")
peripheral_table_columns.set_unit("column_01", "column_01")
population_placeholder_columns = population_table_columns.to_placeholder()
peripheral_placeholder_columns = peripheral_table_columns.to_placeholder()
population_placeholder_columns.join(peripheral_placeholder_columns,
"join_key", "time_stamp")
# ----------------------------------------------------------------
model_columns, features_columns, yhat_columns, scores_columns = _fit_model(
population_table_columns, peripheral_table_columns,
population_placeholder_columns, peripheral_placeholder_columns, seed,
dict())
# ----------------------------------------------------------------
# Assign units to Model
population_table_model, peripheral_table_model = datasets.make_same_units_numerical(
random_state=seed)
population_placeholder_model = population_table_model.to_placeholder()
peripheral_placeholder_model = peripheral_table_model.to_placeholder()
population_placeholder_model.join(peripheral_placeholder_model, "join_key",
"time_stamp")
# ----------------------------------------------------------------
model_model, features_model, yhat_model, scores_model = _fit_model(
population_table_model, peripheral_table_model,
population_placeholder_model, peripheral_placeholder_model, seed,
units)
# ----------------------------------------------------------------
# Check whether the results are the same.
assert scores_model == scores_columns
assert (yhat_model == yhat_columns).all()
# ----------------------------------------------------------------
engine.delete_project("examples")
port=3306,
dbname="mydb",
user="******",
password="******",
time_formats=['%Y/%m/%d']
)
# -----------------------------------------------------------------------------
# Set up folders - you need to insert folders on your computer
# The folder that contains expd151.csv
RAW_DATA_FOLDER = os.getenv("HOME") + "/Downloads/diary15/"
# -----------------------------------------------------------------------------
engine.set_project("CE")
# -----------------------------------------------------------------------------
# Begin timing
begin = time.time()
# #############################################################################
# Load data.
# -----------------------------------------------------------------------------
# Load EXPD
expd_fnames = [
RAW_DATA_FOLDER + "expd151.csv",
RAW_DATA_FOLDER + "expd152.csv",
# ```
#
# ## Staging the data
#
# We will start by starting a new project in the getML engine and loading
# the prepared data tables into the Python environment.
# In[ ]:
import os
import pandas as pd
import getml.engine as engine
engine.set_project("gettingStarted")
# Location inside this repository the data is kept.
source_path = os.path.join(os.getcwd(), "../../../data/consumer_expenditures/")
CE_population_training = pd.read_csv(os.path.join(source_path, "CE_population_training.csv"))
CE_population_validation = pd.read_csv(os.path.join(source_path, "CE_population_validation.csv"))
CE_peripheral = pd.read_csv(os.path.join(source_path, "CE_peripheral.csv"))
# In order for the automated feature engineering to get the most out of
# the data, we have to provided some additional information about its
# content. If a column contains e.g. the type of a product encoded in
# integers, operations like comparisons, summation, or the extraction
# the maximum would most probably make no sense. It, therefore, needs to
# be of recognized as *categorical* instead of *discrete*.
