def get_changepoint_BinCheb(self, instances, is_normal=False):
cheb = ""
if self.is_k:
cheb = chebyshev_base(min_size=self.min_winsize,
max_window=self.max_winsize,
k=self.K)
else:
cheb = chebyshev_base_e(min_size=self.min_winsize,
max_window=self.max_winsize,
e=self.E)
bin = binning(bin_period=self.Bin_size)
change_points = []
temp_bin = []
mean_list = []
# instances = data.get_dataset_test(i)
for index, instance in enumerate(instances):
is_change = False
can_add_bin = bin.can_add_bin()
if not can_add_bin:
temp_bin.append(bin.get_bin())
mean_bin = mean(bin.get_bin())
mean_list.append(mean_bin)
is_change = cheb.add_element(mean_bin)
bin = binning(bin_period=self.Bin_size)
bin.add_instance(instance)
if is_change:
if is_normal:
for j in range(index - self.Bin_size, index):
change_points.append(j)
else:
change_points.append(index)
self.cheb = cheb
self.ex = cheb.get_exception()
return change_points, mean_list
def get_Changepoint_fix_Bin(bin_size, instances, e, max_window=100):
bin = binning(bin_period=bin_size)
cheb = cheb_noDY(bin_size=bin_size, e=e, max_window=max_window)
change_points = []
for index, instance in enumerate(instances):
bin.add_instance(instance)
can_add_bin = bin.can_add_bin()
if not can_add_bin:
mean_bin = mean(bin.get_bin())
bln_change = cheb.add_element(mean_bin)
if bln_change:
change_points.append(index)
bin = binning(bin_period=bin_size)
return change_points
def cal_k_tran(data,input_k,input_bin):
data.clear_change_points()
cheb = chebyshev_base(min_size=60, max_window=60, k=input_k)
bin = binning(bin_period=input_bin)
for i, file_name in enumerate(data.get_files_test()):
change_points = []
instances = data.get_dataset_test(i)
for index, instance in enumerate(instances):
is_change = False
can_add_bin = bin.can_add_bin()
if can_add_bin:
bin.add_instance(instance)
else:
mean_bin = mean(bin.get_bin())
is_change = cheb.add_element(mean_bin)
bin.reset_bin()
# if is_change:
# change_points.append(index)
if is_change:
for j in range(index - bin_period + 1, index + 1):
change_points.append(j)
data.set_change_points(change_points)
return data
def gen_dataBin(instances, input_bin):
mean_list = []
bin = binning(bin_period=input_bin)
for index_in , instance in enumerate(instances):
can_add_bin = bin.can_add_bin()
if not can_add_bin:
mean_bin = bin.get_mean_bin()
for i in range(input_bin):
mean_list.append(mean_bin)
bin = binning(bin_period=input_bin)
bin.add_instance(instance)
if bin.current_size()>0:
mean_bin = bin.get_mean_bin()
for i in range(bin.current_size()):
mean_list.append(mean_bin)
return mean_list
def convert_databin(bin_size, instances):
bin = binning(bin_period=bin_size)
data_after_bin = []
for index, instance in enumerate(instances):
can_add_bin = bin.can_add_bin()
if not can_add_bin:
mean_bin = mean(bin.get_bin())
for i in range(bin_size):
data_after_bin.append(mean_bin)
bin = binning(bin_period=bin_size)
bin.add_instance(instance)
if index == len(instances) - 1:
if len(bin.get_bin()) >= 0:
mean_bin = mean(bin.get_bin())
for i in range(len(bin.get_bin())):
data_after_bin.append(mean_bin)
return data_after_bin
def get_changepoint_skmethod(self, instances, is_normal=False):
change_points = []
ks = sk_detection(k=self.K)
bin = binning(bin_period=self.Bin_size)
for index, instance in enumerate(instances):
is_change = False
can_add_bin = bin.can_add_bin()
if not can_add_bin:
is_change = ks.add_element(bin)
bin = binning(bin_period=self.Bin_size)
bin.add_instance(instance)
if is_change:
if is_normal:
for j in range(index - self.Bin_size, index):
change_points.append(j)
else:
change_points.append(index)
return change_points
def __init__(self, max_window=100, k=3, bin_size=3):
super().__init__()
self.window = []
self.sample_count = None
self._width = 0
self.k = k
self._variance = None
self.max_size = max_window
self.count_detect = 0
self.reset()
self.mean_list = []
self.std_low_list = []
self.std_high_list = []
self.bin_size = bin_size
self.bin = binning(bin_period=bin_size)
def reset(self):
""" reset
Resets the change detector parameters.
"""
super().reset()
self.sample_count = 1
# self.mean = 1.0
self._variance = 0.0
self._width = 0
self.window = []
self.mean_list = []
self.std_low_list = []
self.std_high_list = []
self.bin_size = 3
self.bin = binning(bin_period=self.bin_size)
def __init__(self, max_window=100, k=3, bin_size=3):
super().__init__()
self.window = []
self.sample_count = None
self._width = 0
self.k = k
self._total = None
# self._mean = None
self._variance = None
self.max_size = max_window
# self.warning_level = warning_level
# self.out_control_level = out_control_level
self.count_detect = 0
self.mean_list = []
self.std_low_list = []
self.std_high_list = []
self.bin_size = bin_size
self.bin = binning(bin_period=bin_size)
def get_changepoint_kalmanBinCheb(self,
instances,
inpput_x=10,
input_var=10000,
model_mean=0,
model_variance=0.01):
x_start = inpput_x
x_var_start = input_var
model_mean = model_mean
model_variance = model_variance
cheb = chebyshev_base(min_size=self.min_winsize,
max_window=self.max_winsize,
k=self.K)
bin = binning(bin_period=self.Bin_size)
change_points = []
measurements = instances
# xs = np.zeros((len(measurements), 2))
# ps = np.zeros((len(measurements), 2))
x = x_start
x_var = x_var_start
for index, measurement in enumerate(measurements):
x, x_var = kalman.update(x, x_var, measurement, model_variance)
# xs[i] = x, x_var
x, x_var = kalman.predict(x, x_var, model_mean, model_variance)
# ps[i] = x, x_var
is_change = False
can_add_bin = bin.can_add_bin()
if not can_add_bin:
mean_bin = mean(bin.get_bin())
is_change = cheb.add_element(mean_bin)
bin.reset_bin()
bin.add_instance(x)
if is_change:
for j in range(index - self.bin_period + 1, index + 1):
change_points.append(j)
return change_points
def add_element(self, value):
bln_change = False
self.bin.add_instance(value)
can_add_bin = self.bin.can_add_bin()
if not can_add_bin:
if len(self.window) > 0:
last_bin = self.window[-1]
mean_pre = last_bin.get_mean_bin()
std_pre = last_bin.get_std_bin()
mean_cur = self.bin.get_mean_bin()
std_cur = self.bin.get_std_bin()
kstar = abs(mean_pre - mean_cur) / (std_pre + std_cur)
if kstar >= self.k:
bln_change = True
self.window.append(self.bin)
self.in_concept_change = bln_change
self.bin = binning(bin_period=self.bin_size)
if len(self.window) > self.max_size:
self.window.pop(0)
return bln_change
def cal_k_tran(data, input_k, input_bin):
x_start = 10
x_var_start = 10000
model_mean = 0
model_variance = 0.01
data.clear_change_points()
cheb = chebyshev_base(min_size=60, max_window=60, k=input_k)
bin = binning(bin_period=input_bin)
for i, file_name in enumerate(data.get_files_test()):
change_points = []
measurements = data.get_dataset_test(i)
xs = np.zeros((len(measurements), 2))
ps = np.zeros((len(measurements), 2))
x = x_start
x_var = x_var_start
for index, measurement in enumerate(measurements):
x, x_var = kalman.update(x, x_var, measurement, model_variance)
xs[i] = x, x_var
x, x_var = kalman.predict(x, x_var, model_mean, model_variance)
ps[i] = x, x_var
is_change = False
can_add_bin = bin.can_add_bin()
if not can_add_bin:
mean_bin = mean(bin.get_bin())
is_change = cheb.add_element(mean_bin)
bin.reset_bin()
bin.add_instance(x)
if is_change:
for j in range(index - bin_period + 1, index + 1):
change_points.append(j)
data.set_change_points(change_points)
return data
# interval=I)
data = data_stream(path='D:\\git_project\\data stream\\lightcurve_benchmark\\{}'.format(data_type), type=type,
pattern=pattern, len=L,
interval=I)
data.load_data_fromfile()
# binning process
cheb = chebyshev_base(min_size=60,max_window=60,k=k)
for i,file_name in enumerate(data.get_files_test()):
change_points = []
measurements = data.get_dataset_test(i)
xs = np.zeros((len(measurements), 2))
ps = np.zeros((len(measurements), 2))
bin = binning(bin_period=bin_period)
x = x_start
x_var = x_var_start
for index, measurement in enumerate(measurements):
x, x_var = kalman.update(x, x_var, measurement, model_variance)
xs[i] = x, x_var
x, x_var = kalman.predict(x, x_var, model_mean, model_variance)
ps[i] = x, x_var
is_change = False
can_add_bin = bin.can_add_bin()
if can_add_bin:
bin.add_instance(x)
else:
mean_bin = mean(bin.get_bin())
def resetTempBin(self):
self.bin = binning(bin_period=self.binSize)
def reset(self):
super().reset()
self.bin = binning(bin_period=self.binSize)
def __init__(self, initialBin=3, windowSize=100):
super().__init__(initialBin, windowSize)
self.bin = binning(bin_period=self.binSize)
