def arundo_adtk(data_path):
data = get_data(data_path)
data['date'] = data['timestamp'].apply(
lambda i: datetime.fromtimestamp(i)) # 时间转换
s_train = data[['date', 'value']]
# 设置索引项
s_train = s_train.set_index('date')
s_train = validate_series(s_train)
print(s_train)
# plot(s_train)
# STL分解+离群点检测
steps = [("deseasonal", STLDecomposition(freq=20)),
("quantile_ad", QuantileAD(high=0.9997, low=0.005))]
pipeline = Pipeline(steps)
anomalies = pipeline.fit_detect(s_train)
print(anomalies)
# plot(s_train, anomaly_pred=anomalies, ap_color='red', ap_marker_on_curve=True)
# 绘制检测结果]
known_anomalies = data.loc[data['label'] == 1]
known_anomalies = known_anomalies[['date', 'label']]
known_anomalies = known_anomalies.set_index('date')
known_anomalies = to_events(known_anomalies)
print(known_anomalies)
plot(s_train,
anomaly_true=known_anomalies,
anomaly_pred=anomalies,
ap_color='red',
ap_marker_on_curve=True,
at_color="orange")
plt.savefig(img_path + "arundo_adtk.png", dpi=1000)
plt.show()
def test_pipeline():
s = pd.Series(
[0, 1, 2, 3, 2, 1] * 10,
index=pd.date_range(start="2017-1-1", periods=60, freq="D"),
)
my_pipe = Pipeline([
(
"deseasonal_residual",
transformer.NaiveSeasonalDecomposition(freq=6),
),
(
"abs_residual",
transformer.CustomizedTransformer1D(transform_func=abs),
),
("iqr_ad", detector.InterQuartileRangeAD(c=(None, 3))),
])
my_pipe.fit_detect(s)
def test_transformer_pipe():
s = pd.Series(
[0, 1, 2, 3, 2, 1] * 10,
index=pd.date_range(start="2017-1-1", periods=60, freq="D"),
)
my_pipe = Pipenet(
{
"deseasonal_residual": {
"model": transformer.ClassicSeasonalDecomposition(freq=6),
"input": "original",
},
"abs_residual": {
"model": transformer.CustomizedTransformer1D(
transform_func=abs
),
"input": "deseasonal_residual",
},
}
)
with pytest.raises(RuntimeError, match="`fit_transform`"):
my_pipe.fit_detect(s)
my_pipe.fit(s)
my_pipe.transform(s)
my_pipe.fit_transform(s)
my_pipe = Pipeline(
[
(
"deseasonal_residual",
transformer.ClassicSeasonalDecomposition(freq=6),
),
(
"abs_residual",
transformer.CustomizedTransformer1D(transform_func=abs),
),
]
)
with pytest.raises(RuntimeError, match="`fit_transform`"):
my_pipe.fit_detect(s)
my_pipe.fit(s)
my_pipe.transform(s)
my_pipe.fit_transform(s)
print(len(anomalies_1))
##############################################################################################################
##Pipeline Definition to Combine PersistAD Detector with Double Rolling Aggregate Transformer
##Low Tolerance Model
steps = [
(
"DoubleRolling", DoubleRollingAggregate(agg="median", window=(4, 1))
), ##Calcuates the Difference of Aggregated Metrics Between Two Sliding Windows.
(
"persist", PersistAD(agg="median", side="both", c=4)
), #c is Factor Used to Determine the Bound of Normal Range Based on Historical Interquartile Range. Default Value is: 3.0.
]
pipeline = Pipeline(steps)
##############################################################################################################
##Training Phase
anomalies_2 = pipeline.fit_detect(
s_train, return_list=True
) ##fit_detect Function Trains the Model and Detects Anomalies in Training Set
plot(s_train,
anomaly=anomalies_2,
ts_markersize=1,
anomaly_markersize=2,
anomaly_tag="marker",
anomaly_color='red')
for i in anomalies_2:
def test_pipeline():
s = pd.Series(
[0, 1, 2, 3, 2, 1] * 5 + [0, 1, 2, 4, 2, 1] + [0, 1, 2, 3, 2, 1] * 4,
index=pd.date_range(start="2017-1-1", periods=60, freq="D"),
)
my_pipe = Pipeline(
[
(
"deseasonal_residual",
transformer.ClassicSeasonalDecomposition(freq=6),
),
(
"abs_residual",
transformer.CustomizedTransformer1D(transform_func=abs),
),
("ad", detector.QuantileAD(high=0.99)),
]
)
my_pipe.fit_detect(s)
assert (
my_pipe.score(
s,
pd.Series([0] * 33 + [1] + [0] * 26, index=s.index),
scoring="recall",
)
== 1
)
assert (
my_pipe.score(
s,
pd.Series([0] * 33 + [1] + [0] * 26, index=s.index),
scoring="precision",
)
== 1
)
assert (
my_pipe.score(
s,
pd.Series([0] * 33 + [1] + [0] * 26, index=s.index),
scoring="f1",
)
== 1
)
assert (
my_pipe.score(
s,
pd.Series([0] * 33 + [1] + [0] * 26, index=s.index),
scoring="iou",
)
== 1
)
assert my_pipe.get_params() == {
"deseasonal_residual": {"freq": 6, "trend": False},
"abs_residual": {
"fit_func": None,
"fit_func_params": None,
"transform_func": abs,
"transform_func_params": None,
},
"ad": {"high": 0.99, "low": None},
}