adelorenzo/stable
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #7296 from th0rntwig/dbscan

Browse Source 
Improve MinPts calculation in DBSCAN, add outlier protection, and add data_kitchen tests
This commit is contained in:
Robert Caulk 2022-08-28 14:37:47 +02:00 committed by GitHub
commit 39a739eadb
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 119 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
docs/freqai.md
View File

@ -113,6 +113,7 @@ Mandatory parameters are marked as **Required**, which means that they are requi
| `use_SVM_to_remove_outliers` | Train a support vector machine to detect and remove outliers from the training data set, as well as from incoming data points. See details about how it works [here](#removing-outliers-using-a-support-vector-machine-svm). <br> **Datatype:** Boolean.
| `svm_params` | All parameters available in Sklearn's `SGDOneClassSVM()`. See details about some select parameters [here](#removing-outliers-using-a-support-vector-machine-svm). <br> **Datatype:** Dictionary.
| `use_DBSCAN_to_remove_outliers` | Cluster data using DBSCAN to identify and remove outliers from training and prediction data. See details about how it works [here](#removing-outliers-with-dbscan). <br> **Datatype:** Boolean.
| `outlier_protection_percentage` | If more than `outlier_protection_percentage` fraction of points are removed as outliers, FreqAI will log a warning message and ignore outlier detection while keeping the original dataset intact. <br> **Datatype:** float. Default: `30`
| | **Data split parameters**
| `data_split_parameters` | Include any additional parameters available from Scikit-learn `test_train_split()`, which are shown [here](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html) (external website). <br> **Datatype:** Dictionary.
| `test_size` | Fraction of data that should be used for testing instead of training. <br> **Datatype:** Positive float < 1.

1
freqtrade/freqai/data_drawer.py
View File

@ -566,7 +566,6 @@ class FreqaiDataDrawer:
for training according to user defined train_period_days
metadata: dict = strategy furnished pair metadata
"""
with self.history_lock:
corr_dataframes: Dict[Any, Any] = {}
base_dataframes: Dict[Any, Any] = {}

49
freqtrade/freqai/data_kitchen.py
View File

@ -513,6 +513,19 @@ class FreqaiDataKitchen:
return avg_mean_dist
def get_outlier_percentage(self, dropped_pts: npt.NDArray) -> float:
"""
Check if more than X% of points werer dropped during outlier detection.
"""
outlier_protection_pct = self.freqai_config["feature_parameters"].get(
"outlier_protection_percentage", 30)
outlier_pct = (dropped_pts.sum() / len(dropped_pts)) * 100
if outlier_pct >= outlier_protection_pct:
self.svm_model = None
return outlier_pct
else:
return 0.0
def use_SVM_to_remove_outliers(self, predict: bool) -> None:
"""
Build/inference a Support Vector Machine to detect outliers
@ -550,8 +563,16 @@ class FreqaiDataKitchen:
self.data_dictionary["train_features"]
)
y_pred = self.svm_model.predict(self.data_dictionary["train_features"])
dropped_points = np.where(y_pred == -1, 0, y_pred)
kept_points = np.where(y_pred == -1, 0, y_pred)
# keep_index = np.where(y_pred == 1)
outlier_pct = self.get_outlier_percentage(1 - kept_points)
if outlier_pct:
logger.warning(
f"SVM detected {outlier_pct:.2f}% of the points as outliers. "
f"Keeping original dataset."
)
return
self.data_dictionary["train_features"] = self.data_dictionary["train_features"][
(y_pred == 1)
]
@ -563,7 +584,7 @@ class FreqaiDataKitchen:
]
logger.info(
f"SVM tossed {len(y_pred) - dropped_points.sum()}"
f"SVM tossed {len(y_pred) - kept_points.sum()}"
f" train points from {len(y_pred)} total points."
)
@ -572,7 +593,7 @@ class FreqaiDataKitchen:
# to reduce code duplication
if self.freqai_config['data_split_parameters'].get('test_size', 0.1) != 0:
y_pred = self.svm_model.predict(self.data_dictionary["test_features"])
dropped_points = np.where(y_pred == -1, 0, y_pred)
kept_points = np.where(y_pred == -1, 0, y_pred)
self.data_dictionary["test_features"] = self.data_dictionary["test_features"][
(y_pred == 1)
]
@ -583,7 +604,7 @@ class FreqaiDataKitchen:
]
logger.info(
f"SVM tossed {len(y_pred) - dropped_points.sum()}"
f"SVM tossed {len(y_pred) - kept_points.sum()}"
f" test points from {len(y_pred)} total points."
)
@ -635,8 +656,8 @@ class FreqaiDataKitchen:
cos(angle) * (point[1] - origin[1])
return (x, y)
MinPts = len(self.data_dictionary['train_features'].columns) * 2
# measure pairwise distances to train_features.shape[1]*2 nearest neighbours
MinPts = int(len(self.data_dictionary['train_features'].index) * 0.25)
# measure pairwise distances to nearest neighbours
neighbors = NearestNeighbors(
n_neighbors=MinPts, n_jobs=self.thread_count)
neighbors_fit = neighbors.fit(self.data_dictionary['train_features'])
@ -667,6 +688,14 @@ class FreqaiDataKitchen:
self.data['DBSCAN_min_samples'] = MinPts
dropped_points = np.where(clustering.labels_ == -1, 1, 0)
outlier_pct = self.get_outlier_percentage(dropped_points)
if outlier_pct:
logger.warning(
f"DBSCAN detected {outlier_pct:.2f}% of the points as outliers. "
f"Keeping original dataset."
)
return
self.data_dictionary['train_features'] = self.data_dictionary['train_features'][
(clustering.labels_ != -1)
]
@ -722,6 +751,14 @@ class FreqaiDataKitchen:
0,
)
outlier_pct = self.get_outlier_percentage(1 - do_predict)
if outlier_pct:
logger.warning(
f"DI detected {outlier_pct:.2f}% of the points as outliers. "
f"Keeping original dataset."
)
return
if (len(do_predict) - do_predict.sum()) > 0:
logger.info(
f"DI tossed {len(do_predict) - do_predict.sum()} predictions for "

46
tests/freqai/conftest.py
View File

@ -1,5 +1,6 @@
from copy import deepcopy
from pathlib import Path
from unittest.mock import MagicMock
import pytest
@ -81,6 +82,51 @@ def get_patched_freqaimodel(mocker, freqaiconf):
return freqaimodel
def make_data_dictionary(mocker, freqai_conf):
freqai_conf.update({"timerange": "20180110-20180130"})
strategy = get_patched_freqai_strategy(mocker, freqai_conf)
exchange = get_patched_exchange(mocker, freqai_conf)
strategy.dp = DataProvider(freqai_conf, exchange)
strategy.freqai_info = freqai_conf.get("freqai", {})
freqai = strategy.freqai
freqai.live = True
freqai.dk = FreqaiDataKitchen(freqai_conf)
freqai.dk.pair = "ADA/BTC"
timerange = TimeRange.parse_timerange("20180110-20180130")
freqai.dd.load_all_pair_histories(timerange, freqai.dk)
freqai.dd.pair_dict = MagicMock()
data_load_timerange = TimeRange.parse_timerange("20180110-20180130")
new_timerange = TimeRange.parse_timerange("20180120-20180130")
corr_dataframes, base_dataframes = freqai.dd.get_base_and_corr_dataframes(
data_load_timerange, freqai.dk.pair, freqai.dk
)
unfiltered_dataframe = freqai.dk.use_strategy_to_populate_indicators(
strategy, corr_dataframes, base_dataframes, freqai.dk.pair
)
unfiltered_dataframe = freqai.dk.slice_dataframe(new_timerange, unfiltered_dataframe)
freqai.dk.find_features(unfiltered_dataframe)
features_filtered, labels_filtered = freqai.dk.filter_features(
unfiltered_dataframe,
freqai.dk.training_features_list,
freqai.dk.label_list,
training_filter=True,
)
data_dictionary = freqai.dk.make_train_test_datasets(features_filtered, labels_filtered)
data_dictionary = freqai.dk.normalize_data(data_dictionary)
return freqai
def get_freqai_live_analyzed_dataframe(mocker, freqaiconf):
strategy = get_patched_freqai_strategy(mocker, freqaiconf)
exchange = get_patched_exchange(mocker, freqaiconf)

30
tests/freqai/test_freqai_datakitchen.py
View File

@ -5,7 +5,8 @@ from pathlib import Path
import pytest
from freqtrade.exceptions import OperationalException
from tests.freqai.conftest import get_patched_data_kitchen
from tests.conftest import log_has_re
from tests.freqai.conftest import get_patched_data_kitchen, make_data_dictionary
@pytest.mark.parametrize(
@ -66,3 +67,30 @@ def test_check_if_model_expired(mocker, freqai_conf, timestamp, expected):
dk = get_patched_data_kitchen(mocker, freqai_conf)
assert dk.check_if_model_expired(timestamp) == expected
shutil.rmtree(Path(dk.full_path))
def test_use_DBSCAN_to_remove_outliers(mocker, freqai_conf, caplog):
freqai = make_data_dictionary(mocker, freqai_conf)
# freqai_conf['freqai']['feature_parameters'].update({"outlier_protection_percentage": 1})
freqai.dk.use_DBSCAN_to_remove_outliers(predict=False)
assert log_has_re(
"DBSCAN found eps of 2.42.",
caplog,
)
def test_compute_distances(mocker, freqai_conf):
freqai = make_data_dictionary(mocker, freqai_conf)
freqai_conf['freqai']['feature_parameters'].update({"DI_threshold": 1})
avg_mean_dist = freqai.dk.compute_distances()
assert round(avg_mean_dist, 2) == 2.56
def test_use_SVM_to_remove_outliers_and_outlier_protection(mocker, freqai_conf, caplog):
freqai = make_data_dictionary(mocker, freqai_conf)
freqai_conf['freqai']['feature_parameters'].update({"outlier_protection_percentage": 0.1})
freqai.dk.use_SVM_to_remove_outliers(predict=False)
assert log_has_re(
"SVM detected 8.46%",
caplog,
)