Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add ARIMA Normalization Functionality #89

Open
wants to merge 10 commits into
base: master
Choose a base branch
from
Open

Add ARIMA Normalization Functionality #89

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
409e5ba
add arima functionality
ml874 Jan 29, 2020
e6ab976
fixed arima functionality
ml874 Feb 13, 2020
a577fa5
added moving average component to arima
ml874 Feb 26, 2020
5967956
additional changes, pep8 formatting
ml874 Mar 18, 2020
d9918e0
additional changes, modified test file
ml874 Apr 2, 2020
68d3cc2
moved arima normalization to econometrics.py
ml874 Apr 7, 2020
36d3294
remove arima from timeseries __init__.py
ml874 Apr 8, 2020
94a1596
pep8 and added code coverage for econometrics
ml874 Apr 20, 2020
e835ddc
pep8 changes
ml874 Apr 20, 2020
6b9cc40
more code coverage
ml874 Apr 20, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
157 changes: 157 additions & 0 deletions gs_quant/test/timeseries/test_econometrics.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,10 @@
under the License.
"""

from math import isclose

import pytest
from pandas import Timestamp
from pandas.util.testing import assert_series_equal

from gs_quant.timeseries import *
Expand Down Expand Up @@ -366,5 +369,159 @@ def test_max_drawdown():
assert_series_equal(output_window, pd.Series([0.0, 0.0, 0.0, -0.2, -0.2, -0.75]), obj="Max drawdown window")


def test_arima_fit():
test_dict = {
'High': {
Timestamp('1989-01-03 00:00:00'): 3.575721263885498,
Timestamp('1989-01-04 00:00:00'): 3.5857372283935547,
Timestamp('1989-01-05 00:00:00'): 3.62580132484436,
Timestamp('1989-01-06 00:00:00'): 3.62580132484436,
Timestamp('1989-01-09 00:00:00'): 3.575721263885498,
Timestamp('1989-01-10 00:00:00'): 3.575721263885498,
Timestamp('1989-01-11 00:00:00'): 3.5657050609588623,
Timestamp('1989-01-12 00:00:00'): 3.635817289352417,
Timestamp('1989-01-13 00:00:00'): 3.615785360336304,
Timestamp('1989-01-16 00:00:00'): 3.615785360336304,
Timestamp('1989-01-17 00:00:00'): 3.635817289352417,
Timestamp('1989-01-18 00:00:00'): 3.675881385803223,
Timestamp('1989-01-19 00:00:00'): 3.695913553237915,
Timestamp('1989-01-20 00:00:00'): 3.665865421295166,
Timestamp('1989-01-23 00:00:00'): 3.675881385803223,
Timestamp('1989-01-24 00:00:00'): 3.675881385803223,
Timestamp('1989-01-25 00:00:00'): 3.695913553237915,
Timestamp('1989-01-26 00:00:00'): 3.7760417461395264,
Timestamp('1989-01-27 00:00:00'): 3.8561699390411377,
Timestamp('1989-01-30 00:00:00'): 3.8561699390411377},
'Low': {
Timestamp('1989-01-03 00:00:00'): 3.4855768680572514,
Timestamp('1989-01-04 00:00:00'): 3.5356571674346924,
Timestamp('1989-01-05 00:00:00'): 3.575721263885498,
Timestamp('1989-01-06 00:00:00'): 3.575721263885498,
Timestamp('1989-01-09 00:00:00'): 3.5356571674346924,
Timestamp('1989-01-10 00:00:00'): 3.5356571674346924,
Timestamp('1989-01-11 00:00:00'): 3.5256409645080566,
Timestamp('1989-01-12 00:00:00'): 3.5456731319427486,
Timestamp('1989-01-13 00:00:00'): 3.5857372283935547,
Timestamp('1989-01-16 00:00:00'): 3.5957531929016118,
Timestamp('1989-01-17 00:00:00'): 3.5857372283935547,
Timestamp('1989-01-18 00:00:00'): 3.615785360336304,
Timestamp('1989-01-19 00:00:00'): 3.655849456787109,
Timestamp('1989-01-20 00:00:00'): 3.62580132484436,
Timestamp('1989-01-23 00:00:00'): 3.615785360336304,
Timestamp('1989-01-24 00:00:00'): 3.615785360336304,
Timestamp('1989-01-25 00:00:00'): 3.655849456787109,
Timestamp('1989-01-26 00:00:00'): 3.665865421295166,
Timestamp('1989-01-27 00:00:00'): 3.79607367515564,
Timestamp('1989-01-30 00:00:00'): 3.786057710647583},
'Close': {
Timestamp('1989-01-03 00:00:00'): 3.5256409645080566,
Timestamp('1989-01-04 00:00:00'): 3.5857372283935547,
Timestamp('1989-01-05 00:00:00'): 3.575721263885498,
Timestamp('1989-01-06 00:00:00'): 3.575721263885498,
Timestamp('1989-01-09 00:00:00'): 3.575721263885498,
Timestamp('1989-01-10 00:00:00'): 3.5556890964508057,
Timestamp('1989-01-11 00:00:00'): 3.5556890964508057,
Timestamp('1989-01-12 00:00:00'): 3.605769157409668,
Timestamp('1989-01-13 00:00:00'): 3.605769157409668,
Timestamp('1989-01-16 00:00:00'): 3.5957531929016118,
Timestamp('1989-01-17 00:00:00'): 3.62580132484436,
Timestamp('1989-01-18 00:00:00'): 3.675881385803223,
Timestamp('1989-01-19 00:00:00'): 3.665865421295166,
Timestamp('1989-01-20 00:00:00'): 3.6458332538604736,
Timestamp('1989-01-23 00:00:00'): 3.62580132484436,
Timestamp('1989-01-24 00:00:00'): 3.675881385803223,
Timestamp('1989-01-25 00:00:00'): 3.675881385803223,
Timestamp('1989-01-26 00:00:00'): 3.756009578704834,
Timestamp('1989-01-27 00:00:00'): 3.79607367515564,
Timestamp('1989-01-30 00:00:00'): 3.846153736114502},
}

test_df = pd.DataFrame(test_dict)
arima = econometrics.arima()

train_size_values = [0.75, int(0.75*len(test_df)), None]
for train_size in train_size_values:
arima.fit(test_df, train_size=train_size, freq='B', q_vals=[0])
transformed_test_df = arima.transform(test_df)

for col in transformed_test_df.keys():
count_nans = arima.best_params[col].p + arima.best_params[col].d
assert(count_nans == transformed_test_df[col].isna().sum())

# Test (2,1,0) Model
test_df_high = test_df['High'].diff()
assert(isclose(transformed_test_df['High'][3], (arima.best_params['High'].const + test_df_high[2] *
arima.best_params['High'].ar_coef[0] + test_df_high[1] *
arima.best_params['High'].ar_coef[1]), abs_tol=1e-8))
assert(isclose(transformed_test_df['High'][4], (arima.best_params['High'].const + test_df_high[3] *
arima.best_params['High'].ar_coef[0] + test_df_high[2] *
arima.best_params['High'].ar_coef[1]), abs_tol=1e-8))
assert(isclose(transformed_test_df['High'][-1], (arima.best_params['High'].const + test_df_high[-2] *
arima.best_params['High'].ar_coef[0] + test_df_high[-3] *
arima.best_params['High'].ar_coef[1]), abs_tol=1e-8))

# Test (2,2,0) Model
test_df_low = test_df['Low'].diff().diff()
assert(isclose(transformed_test_df['Low'][4], (arima.best_params['Low'].const + test_df_low[3] *
arima.best_params['Low'].ar_coef[0] + test_df_low[2] *
arima.best_params['Low'].ar_coef[1]), abs_tol=1e-8))
assert(isclose(transformed_test_df['Low'][5], (arima.best_params['Low'].const + test_df_low[4] *
arima.best_params['Low'].ar_coef[0] + test_df_low[3] *
arima.best_params['Low'].ar_coef[1]), abs_tol=1e-8))
assert(isclose(transformed_test_df['Low'][-1], (arima.best_params['Low'].const + test_df_low[-2] *
arima.best_params['Low'].ar_coef[0] + test_df_low[-3] *
arima.best_params['Low'].ar_coef[1]), abs_tol=1e-8))

# Test (2,1,0) Model
test_df_close = test_df['Close'].diff()
assert(isclose(transformed_test_df['Close'][3], (arima.best_params['Close'].const + test_df_close[2] *
arima.best_params['Close'].ar_coef[0] + test_df_close[1] *
arima.best_params['Close'].ar_coef[1]), abs_tol=1e-8))
assert(isclose(transformed_test_df['Close'][4], (arima.best_params['Close'].const + test_df_close[3] *
arima.best_params['Close'].ar_coef[0] + test_df_close[2] *
arima.best_params['Close'].ar_coef[1]), abs_tol=1e-8))
assert(isclose(transformed_test_df['Close'][-1], (arima.best_params['Close'].const + test_df_close[-2] *
arima.best_params['Close'].ar_coef[0] + test_df_close[-3] *
arima.best_params['Close'].ar_coef[1]), abs_tol=1e-8))

# Test if input is pd.Series
test_high_series = pd.Series(test_df['High'])
arima.fit(test_high_series, train_size=0.75, freq='B', q_vals=[0])
transformed_test_series = arima.transform(test_high_series)
test_series_high = test_df['High'].diff()
assert(isclose(transformed_test_series['High'][3], (arima.best_params['High'].const + test_series_high[2] *
arima.best_params['High'].ar_coef[0] + test_series_high[1] *
arima.best_params['High'].ar_coef[1]), abs_tol=1e-8))
assert(isclose(transformed_test_series['High'][4], (arima.best_params['High'].const + test_series_high[3] *
arima.best_params['High'].ar_coef[0] + test_series_high[2] *
arima.best_params['High'].ar_coef[1]), abs_tol=1e-8))
assert(isclose(transformed_test_series['High'][-1], (arima.best_params['High'].const + test_series_high[-2] *
arima.best_params['High'].ar_coef[0] + test_series_high[-3] *
arima.best_params['High'].ar_coef[1]), abs_tol=1e-8))

# Test if p=0 and d=0
new_arima = econometrics.arima()
zero_resid = test_high_series.copy(deep=True)
zero_resid[:] = 0
new_arima.best_params = {'High': econometrics.ARIMA_BestParams(p=0, q=0, d=0,
const=0, ar_coef=[0], ma_coef=[], resid=zero_resid)}

transformed_test_df = new_arima.transform(test_high_series)
assert_series_equal(transformed_test_df['High'], test_df['High'])

# Test if train_size is str
with pytest.raises(ValueError, match='train_size is not int, float, or None'):
arima.fit(test_df, train_size='str', freq='B', q_vals=[0])

# Test if input is list
with pytest.raises(ValueError, match='Not DataFrame or Series!'):
arima.fit([1, 2, 3, 4], train_size=0.75, freq='B', q_vals=[0])

# Test transform with list
with pytest.raises(ValueError, match='Not DataFrame or Series!'):
arima.fit(test_df, train_size=train_size, freq='B', q_vals=[0])
transformed_test_df = arima.transform([1, 2, 3, 4])


if __name__ == "__main__":
pytest.main(args=["test_econometrics.py"])
Loading