def detectChangePoints(self, rawData, filename):
detector = cpdetect.cpDetector(rawData,
distribution='normal',
log_odds_threshold=1)
detector.detect_cp()
detector.to_csv(filename)
return self.csvExtractCps(filename)
from cpdetect import cpDetector
from cpdetect.cp_detector import (Normal, LogNormal)
from cpdetect.tests.utils import get_fn
import numpy as np
import unittest
import pandas as pd
from scipy.special import gammaln
data = np.load(get_fn('data.npy'))
data_2 = np.load(get_fn('data2.npz'))
# Convert to cpDetect format
data_2 = [data_2[i] for i in data_2.files]
detector = cpDetector(data_2,
distribution='log_normal',
log_odds_threshold=-10)
class TestCpDetect(unittest.TestCase):
def test_lognormal(self):
""" Test Log-normal mean and variance """
mean, var = LogNormal.mean_var(data)
self.assertEqual(mean, 1.0081320131891722)
self.assertEqual(var, 0.010999447786412363)
def test_normal(self):
"""Test Normal mean and variance"""
mean, var = Normal.mean_var(data)
import cPickle as pickle
except:
import pickle
import matplotlib.pyplot as plt
import matplotlib as mpl
from matplotlib.backends.backend_pdf import PdfPages
# Load synthetic trajectories
trajs = np.load('synthetic_trajs.np.npy')
true_step = pickle.load(open('step_synthetic.pickle', 'rb'))
trajs = trajs.tolist()
# Run them through cpdetect for thresholds (-10, 0)
for i in range(10):
detector = cpDetector(trajs,
distribution='log_normal',
log_odds_threshold=-i)
detector.detect_cp()
# pickle detector
pickle.dump(detector, open('detector_{}.pickle'.format(str(i)), 'wb'))
# save steps
df = pd.DataFrame.from_dict(detector.step_function, orient='index')
df.to_csv('step_function_{}.csv'.format(str(i)))
detector.to_csv('ts_log_odds_{}.csv'.format(str(i)))
# Plot
filename = 'synthetic_{}.pdf'.format(str(i))
fontsize = 6
x_spacing = 1000
time_res = 1.0
chunk = len(trajs) / 4
true_step = pickle.load(open('step_synthetic.pickle', 'rb'))
# run through filter
windows = [2, 4, 6, 8, 16]
M = 12
p = 30
filtered_trajs = []
for traj in trajs:
filtered_trajs.append(nonlinear_filter.nfl_filter(traj, windows, M, p))
# save filtered trajs
np.save(file='filtered_trajs', arr=filtered_trajs)
# Run them through cpdetect for thresholds (-10, 0)
for i in range(11):
detector = cpDetector(filtered_trajs, distribution='log_normal', log_odds_threshold=-i)
detector.detect_cp()
# pickle detector
pickle.dump(detector, open('filtered_detector_{}.pickle'.format(str(i)), 'wb'))
# save steps
df = pd.DataFrame.from_dict(detector.step_function, orient='index')
df.to_csv('filtered_step_function_{}.csv'.format(str(i)))
detector.to_csv('ts_log_odds_filtered_{}.csv'.format(str(i)))
# Plot
filename = 'synthetic_filtered_{}.pdf'.format(str(i))
fontsize = 6
x_spacing = 1000
time_res = 1.0
chunk = len(trajs)/4
with PdfPages(filename) as pdf:
""" Test Change Point detector """
from cpdetect import cpDetector
from cpdetect.cp_detector import (Normal, LogNormal)
from cpdetect.tests.utils import get_fn
import numpy as np
import unittest
import pandas as pd
from scipy.special import gammaln
data = np.load(get_fn('data.npy'))
data_2 = np.load(get_fn('data2.npz'))
# Convert to cpDetect format
data_2 = [data_2[i] for i in data_2.files]
detector = cpDetector(data_2, distribution='log_normal', log_odds_threshold=-10)
class TestCpDetect(unittest.TestCase):
def test_lognormal(self):
""" Test Log-normal mean and variance """
mean, var = LogNormal.mean_var(data)
self.assertEqual(mean, 1.0081320131891722)
self.assertEqual(var, 0.010999447786412363)
def test_normal(self):
"""Test Normal mean and variance"""
mean, var = Normal.mean_var(data)