import numpy as np
from wildboar.datasets import load_dataset
from wildboar.ensemble import ShapeletForestClassifier
from wildboar.explain.counterfactual import counterfactuals
x_train, x_test, y_train, y_test = load_dataset("TwoLeadECG",
repository="wildboar/ucr",
merge_train_test=False)
# x_train = x_train.repeat(2, axis=0).reshape(x_train.shape[0], 2, -1)
# print(x_train[0])
# x_test = x_test.repeat(2, axis=0).reshape(x_test.shape[0], 2, -1)
clf = ShapeletForestClassifier(metric="euclidean",
random_state=1,
n_jobs=-1,
n_estimators=100)
clf.fit(x_train, y_train)
print(clf.score(x_test, y_test))
y_pred = clf.predict(x_test)
class_ = clf.classes_[1]
print("Class: %s" % class_)
print("Pred: %r" % y_pred)
x_test = x_test[y_pred != class_][:10]
y_test = y_test[y_pred != class_][:10]
x_counterfactual, success, score = counterfactuals(clf,
x_test,
class_,
random_state=123,
scoring="euclidean")
x, y = load_dataset("Car")
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.2,
random_state=random_state)
f_embedding = make_pipeline(
ShapeletForestEmbedding(sparse_output=False, random_state=random_state),
PCA(n_components=2, random_state=random_state),
)
f_embedding.fit(x_train)
x_embedding = f_embedding.transform(x_test)
classifiers = [
("Shapelet forest", ShapeletForestClassifier(random_state=random_state)),
("Nearest neighbors", KNeighborsClassifier()),
]
classes = np.unique(y)
n_classes = len(classes)
fig, ax = plt.subplots(
nrows=len(classifiers),
ncols=n_classes,
figsize=(3 * n_classes, 6),
sharex=True,
sharey=True,
)
for i, (name, clf) in enumerate(classifiers):
clf.fit(x_train, y_train)
import numpy as np
from sklearn.model_selection import cross_val_score
from wildboar import datasets
from wildboar.ensemble import ExtraShapeletTreesClassifier, ShapeletForestClassifier
x, y = datasets.load_gun_point()
extra = ExtraShapeletTreesClassifier(
n_estimators=100, n_jobs=-1, metric="scaled_euclidean"
)
rsf = ShapeletForestClassifier(n_estimators=100, n_jobs=-1, metric="scaled_euclidean")
score = cross_val_score(rsf, x, y, cv=10)
print("RSF", np.mean(score))
score = cross_val_score(extra, x, y, cv=10)
print("Extra", np.mean(score))
# etc...
[[0, 1, 9, 1, 0], [0, 1, 2, 3, 4]],
[[1, 2, 3, 0, 0], [0, 0, 0, 1, 2]],
[[0, 0, -1, 0, 1], [1, 2, 3, 0, 1]],
]
# `x` is an array of shape `[5, 2, 5]`, i.e., 5 examples with 2
# dimensions consisting of 5 timesteps
x = np.array(x, dtype=np.float64)
n_samples, n_dimensions, n_timesteps = x.shape
# `y` is the output target
y = np.array([0, 0, 1, 1, 0])
random_state = np.random.RandomState(123)
order = np.arange(n_samples)
random_state.shuffle(order)
x = x[order, :, :]
y = y[order]
f = ShapeletForestClassifier(random_state=random_state, n_shapelets=1)
c = time.time()
f.fit(x, y)
predict = 1
print("Predicting the class of example:")
print(x[predict, :].reshape(-1, n_dimensions, n_timesteps))
print("The true class is:", y[predict], "and the predicted class is:",
f.predict(x[predict, :].reshape(1, n_dimensions, -1))[0])
from wildboar.ensemble import ShapeletForestClassifier
from wildboar.explain.counterfactual import counterfactuals
random_state = 1234
x, y = load_dataset("GunPoint")
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.2,
random_state=random_state)
classifiers = [
(
"Shapelet Forest Counterfactuals",
ShapeletForestClassifier(metric="euclidean",
random_state=random_state,
n_estimators=100),
),
("KNearest Counterfactuals", KNeighborsClassifier(metric="euclidean")),
("Prototype Counterfactuals",
RandomForestClassifier(random_state=random_state)),
]
fig, ax = plt.subplots(nrows=3, sharex=True)
label = np.unique(y_train)[0]
for i, (name, clf) in enumerate(classifiers):
clf.fit(x_train, y_train)
x_test_sample = x_test[y_test != label]
if isinstance(clf, RandomForestClassifier):
kwargs = {"background_x": x_train, "background_y": y_train}
else:
def counterfactuals_sf():
return counterfactuals_plot(
ShapeletForestClassifier(random_state=123,
n_jobs=-1,
metric="euclidean"))
from sklearn.model_selection import train_test_split
from wildboar.datasets import load_dataset
from wildboar.ensemble import ShapeletForestClassifier
from wildboar.explain import IntervalImportance
x, y = load_dataset("LargeKitchenAppliances")
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.3,
random_state=123)
f = ShapeletForestClassifier(
n_shapelets=1,
n_estimators=100,
n_jobs=-1,
metric="scaled_euclidean",
random_state=123,
)
f.fit(x_train, y_train)
i = IntervalImportance(
n_interval=72,
scoring="accuracy",
domain="frequency",
verbose=True,
random_state=123,
)
i.fit(f, x_test, y_test)
ax = i.plot(
x_test,
def test_shapelet_forest_classifier():
x_train, x_test, y_train, y_test = load_dataset(
"GunPoint", repository="wildboar/ucr-tiny", merge_train_test=False
)
clf = ShapeletForestClassifier(n_estimators=10, n_shapelets=10, random_state=1)
clf.fit(x_train, y_train)
branches = [
(
[1, -1, 3, 4, 5, -1, -1, 8, -1, 10, -1, -1, -1],
[2, -1, 12, 7, 6, -1, -1, 9, -1, 11, -1, -1, -1],
),
(
[1, -1, 3, 4, -1, 6, -1, 8, -1, 10, -1, -1, -1],
[2, -1, 12, 5, -1, 7, -1, 9, -1, 11, -1, -1, -1],
),
(
[1, 2, 3, 4, -1, -1, 7, -1, -1, 10, -1, 12, -1, -1, -1],
[14, 9, 6, 5, -1, -1, 8, -1, -1, 11, -1, 13, -1, -1, -1],
),
(
[1, 2, 3, 4, -1, -1, -1, 8, -1, 10, -1, 12, -1, -1, -1],
[14, 7, 6, 5, -1, -1, -1, 9, -1, 11, -1, 13, -1, -1, -1],
),
(
[1, 2, 3, -1, 5, -1, 7, -1, 9, -1, 11, -1, -1, -1, 15, -1, 17, -1, -1],
[14, 13, 4, -1, 6, -1, 8, -1, 10, -1, 12, -1, -1, -1, 16, -1, 18, -1, -1],
),
(
[1, 2, -1, 4, 5, -1, 7, -1, -1, 10, -1, -1, 13, 14, -1, -1, -1],
[12, 3, -1, 9, 6, -1, 8, -1, -1, 11, -1, -1, 16, 15, -1, -1, -1],
),
(
[1, 2, 3, 4, -1, -1, -1, 8, 9, -1, -1, 12, 13, -1, -1, -1, -1],
[16, 7, 6, 5, -1, -1, -1, 11, 10, -1, -1, 15, 14, -1, -1, -1, -1],
),
(
[1, 2, -1, 4, 5, 6, 7, -1, -1, -1, -1, 12, -1, -1, -1],
[14, 3, -1, 11, 10, 9, 8, -1, -1, -1, -1, 13, -1, -1, -1],
),
(
[1, 2, 3, -1, 5, -1, -1, 8, 9, 10, -1, 12, -1, -1, -1, -1, -1],
[16, 7, 4, -1, 6, -1, -1, 15, 14, 11, -1, 13, -1, -1, -1, -1, -1],
),
(
[1, 2, -1, 4, -1, 6, -1, 8, -1, 10, -1, 12, -1, -1, -1],
[14, 3, -1, 5, -1, 7, -1, 9, -1, 11, -1, 13, -1, -1, -1],
),
]
thresholds = [
(
[
3.728410228070656,
11.127575591141072,
7.383224794807461,
7.109350684315213,
0.9248183559076002,
6.08675185469423,
],
[
3.728410228070656,
11.127575591141072,
7.383224794807461,
7.109350684315213,
0.9248183559076002,
6.08675185469423,
],
),
(
[
2.468504005311855,
7.912505524900922,
1.0551252327034113,
0.8574299925766751,
0.5760307808209804,
0.009237440363224308,
],
[
2.468504005311855,
7.912505524900922,
1.0551252327034113,
0.8574299925766751,
0.5760307808209804,
0.009237440363224308,
],
),
(
[
3.909569808800988,
2.821010442496668,
1.8694668182965288,
0.034583372931197384,
0.8137102058624538,
0.7560554810866997,
2.713102595233928,
],
[
3.909569808800988,
2.821010442496668,
1.8694668182965288,
0.034583372931197384,
0.8137102058624538,
0.7560554810866997,
2.713102595233928,
],
),
(
[
5.391042553752862,
4.420547070721347,
2.2716225008196576,
0.6679258993537478,
1.5471177855226528,
1.2706259403508802,
6.379381672446367,
],
[
5.391042553752862,
4.420547070721347,
2.2716225008196576,
0.6679258993537478,
1.5471177855226528,
1.2706259403508802,
6.379381672446367,
],
),
(
[
2.784221806516613,
3.9613021926565697,
0.43050821107331483,
1.3603965501478146,
1.9817847740610532,
0.557171910946499,
0.023161212907754903,
3.2040403820972045,
0.25123702588573155,
],
[
2.784221806516613,
3.9613021926565697,
0.43050821107331483,
1.3603965501478146,
1.9817847740610532,
0.557171910946499,
0.023161212907754903,
3.2040403820972045,
0.25123702588573155,
],
),
(
[
9.06314095909644,
0.9301861459984877,
1.2749535932250209,
0.6602701901531287,
0.3105779260645574,
3.199344210068309,
1.7444498163002922,
0.9679068532147111,
],
[
9.06314095909644,
0.9301861459984877,
1.2749535932250209,
0.6602701901531287,
0.3105779260645574,
3.199344210068309,
1.7444498163002922,
0.9679068532147111,
],
),
(
[
10.684770463276237,
1.0443634502866903,
2.657944200018761,
0.31997645008775166,
8.506009151805937,
2.5790890876760417,
2.444351040739898,
0.8797498982567451,
],
[
10.684770463276237,
1.0443634502866903,
2.657944200018761,
0.31997645008775166,
8.506009151805937,
2.5790890876760417,
2.444351040739898,
0.8797498982567451,
],
),
(
[
8.903669489275785,
2.558013265746756,
1.9352062567009694,
0.6160338380839283,
1.1133147922166846,
2.6673841033247827,
0.6693157414483296,
],
[
8.903669489275785,
2.558013265746756,
1.9352062567009694,
0.6160338380839283,
1.1133147922166846,
2.6673841033247827,
0.6693157414483296,
],
),
(
[
2.9771351955856753,
3.4048368843307957,
2.847751510400112,
1.2655496884627422,
4.410184513977114,
2.3116642536119203,
0.5858765536466852,
0.7586458184224343,
],
[
2.9771351955856753,
3.4048368843307957,
2.847751510400112,
1.2655496884627422,
4.410184513977114,
2.3116642536119203,
0.5858765536466852,
0.7586458184224343,
],
),
(
[
6.260659343273105,
0.05120063347084325,
0.678745571123132,
5.913261089713139,
0.25431501853894734,
0.27996560751446015,
0.7309024510514174,
],
[
6.260659343273105,
0.05120063347084325,
0.678745571123132,
5.913261089713139,
0.25431501853894734,
0.27996560751446015,
0.7309024510514174,
],
),
]
for estimator, (left, right), (left_threshold, right_threshold) in zip(
clf.estimators_, branches, thresholds
):
assert_equal(left, estimator.tree_.left)
assert_equal(right, estimator.tree_.right)
assert_almost_equal(
left_threshold, estimator.tree_.threshold[estimator.tree_.left > 0]
)
assert_almost_equal(
right_threshold, estimator.tree_.threshold[estimator.tree_.right > 0]
)
import matplotlib.pylab as plt
import numpy as np
from wildboar.datasets import load_dataset
from wildboar.ensemble import ShapeletForestClassifier
from wildboar.explain.counterfactual import PrototypeCounterfactual
x_train, x_test, y_train, y_test = load_dataset("TwoLeadECG",
repository="wildboar/ucr",
merge_train_test=False)
x_test_original = x_test
y_test_original = y_test
# clf = KNeighborsClassifier(n_neighbors=5, metric="euclidean")
clf = ShapeletForestClassifier(n_estimators=100,
metric="scaled_euclidean",
random_state=10,
n_jobs=-1)
# clf = RandomForestClassifier()
clf.fit(x_train, y_train)
cf = PrototypeCounterfactual(
train_x=x_train,
train_y=y_train,
metric="dtw",
metric_params={"r": 0.1},
method="nearest_shapelet",
method_params={
"min_shapelet_size": 0.1,
"max_shapelet_size": 0.2
},
target=0.70,
from sklearn.neighbors import KNeighborsClassifier
from wildboar.datasets import list_datasets, load_dataset
from wildboar.ensemble import ExtraShapeletTreesClassifier, ShapeletForestClassifier
random_state = 1234
classifiers = {
"Nearest neighbors":
KNeighborsClassifier(
n_neighbors=1,
metric="euclidean",
),
"Shapelet forest":
ShapeletForestClassifier(
n_shapelets=10,
metric="scaled_euclidean",
random_state=random_state,
n_jobs=-1,
),
"Extra shapelet trees":
ExtraShapeletTreesClassifier(
metric="scaled_euclidean",
n_jobs=-1,
random_state=random_state,
),
}
repository = "wildboar/ucr-tiny"
datasets = list_datasets(repository)
df = pd.DataFrame(columns=classifiers.keys(), index=datasets, dtype=float)
for dataset in datasets:
print(dataset)
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from wildboar.datasets import load_dataset
from wildboar.ensemble import ShapeletForestClassifier
from wildboar.explain.counterfactual import counterfactuals
from wildboar.linear_model import RocketClassifier
x, y = load_dataset("GunPoint")
x_train, x_test, y_train, y_test = train_test_split(x, y, random_state=123)
rocket = RocketClassifier(n_kernels=1000, n_jobs=-1, random_state=123)
rocket.fit(x_train, y_train)
print("Rocket score", rocket.score(x_test, y_test))
sf = ShapeletForestClassifier(n_shapelets=10, n_jobs=-1, random_state=123)
sf.fit(x_train, y_train)
print("Shapelet score", sf.score(x_test, y_test))
nearest = KNeighborsClassifier(n_jobs=-1)
nearest.fit(x_train, y_train)
print("Neighbors score", nearest.score(x_test, y_test))
x_test_cls = x_test[y_test == 1.0][:2]
for method in [rocket, sf, nearest]:
print(method)
x_counter, x_success = counterfactuals(
method,
x_test_cls,
2.0,
tree = ShapeletTreeClassifier(random_state=10, metric="scaled_dtw")
tree.fit(x, y, sample_weight=np.ones(x.shape[0]) / x.shape[0])
print_tree(tree.root_node_)
print("Score")
print(tree.score(x, y))
print("score_done")
train = np.loadtxt("data/synthetic_control_TRAIN", delimiter=",")
test = np.loadtxt("data/synthetic_control_TEST", delimiter=",")
y = train[:, 0].astype(np.intp)
x = train[:, 1:].astype(np.float64)
i = np.arange(x.shape[0])
np.random.shuffle(i)
x_test = test[:, 1:].astype(np.float64)
y_test = test[:, 0].astype(np.intp)
f = ShapeletForestClassifier(n_shapelets=1,
metric="scaled_dtw",
metric_params={"r": 0.1})
f.fit(x, y)
c = time.time()
f.fit(x, y)
print(f.classes_)
print("acc:", f.score(x_test, y_test))
print(round(time.time() - c) * 1000)