def test_mov_average_expw(self):
"Test mov_average_expw"
ser_a = ma.array(range(150), dtype=np.float32)
ser_b = ser_a.copy()
# set extreme value to ensure masked values are not impacting result
ser_b[0] = 5e34
ser_b[0] = ma.masked
#
res_a = mf.mov_average_expw(ser_a, 10)
res_b = mf.mov_average_expw(ser_b, 10, tol=1e-6)
assert_almost_equal(res_a[-1], 144.5)
assert_almost_equal(res_a, res_b)
# with specified tolerance value, should be 69 masked values in res_b
assert_equal(np.sum(res_b.mask), 69)
def test_mov_average_expw(self):
"Test mov_average_expw"
ser_a = ma.array(range(150), dtype=np.float32)
ser_b = ser_a.copy()
# set extreme value to ensure masked values are not impacting result
ser_b[0] = 5e34
ser_b[0] = ma.masked
#
res_a = mf.mov_average_expw(ser_a, 10)
res_b = mf.mov_average_expw(ser_b, 10, tol=1e-6)
assert_almost_equal(res_a[-1], 144.5)
assert_almost_equal(res_a, res_b)
# with specified tolerance value, should be 69 masked values in res_b
assert_equal(np.sum(res_b.mask), 69)
def test_mov_average_expw_mask(self):
"Make sure that mov_average_expw doesn't modify the initial mask"
N = 256
series = ts.time_series(np.random.rand(N), start_date=ts.Date("D", "2008-01-01"))
series[96:128] = ma.masked
controlmask = np.zeros(N, dtype=bool)
controlmask[96:128] = True
#
test = mf.mov_average_expw(series, 16)
assert_not_equal(test.mask, series.mask)
assert_equal(series.mask, controlmask)
def test_mov_average_expw_mask(self):
"Make sure that mov_average_expw doesn't modify the initial mask"
N = 256
series = ts.time_series(np.random.rand(N),
start_date=ts.Date('D', '2008-01-01'))
series[96:128] = ma.masked
controlmask = np.zeros(N, dtype=bool)
controlmask[96:128] = True
#
test = mf.mov_average_expw(series, 16)
assert_not_equal(test.mask, series.mask)
assert_equal(series.mask, controlmask)
def make_memory_graph(ma=False):
clf()
values = load("memory.average*.json")
for test in values:
instance_count = len(test['instances'])
t = map(lambda x: x['used'], test['instances'])
def sum_s(*args):
return reduce(lambda x, y: x+int(y), args, 0)
m = min(map(len, t))
for i in t:
del i[m:]
test['results'] = map(sum_s, *t)
print len(test['instances'][0]['used'])
test['x'] = range(0, instance_count*len(test['results'])*1000, instance_count*1000)
linestyles = ['-','--','-.',':']
for i in range(0, len(values)):
x = values[i]['x']
y = values[i]['results']
if ma:
y = mov_average_expw(y, 2000)
gb = 1.0 * 1024**3
y = map(lambda x: x/gb, y)
x = map(lambda x: 1.0*x/1000000, x)
r = plot(x, y, label="%s instances" % (len(values[i]['instances'])), color='k', linestyle=linestyles[i])
legend(loc='upper left')
import numpy as np import matplotlib.pyplot as plt import scikits.timeseries as ts import scikits.timeseries.lib.plotlib as tpl from scikits.timeseries.lib.moving_funcs import mov_average_expw # generate some random data data = np.cumprod(1 + np.random.normal(0, 1, 300) / 100) series = ts.time_series(data, start_date=ts.Date(freq='M', year=1982, month=1)) fig = tpl.tsfigure() fsp = fig.add_tsplot(111) fsp.tsplot(series, '-', mov_average_expw(series, 40), 'r--') plt.show()
import numpy as np from scikits.timeseries import * from scikits.timeseries.lib.moving_funcs import mov_mean, mov_average_expw import utilfuncs.finutils as fn; reload(fn) from matplotlib.pyplot import plot data = np.arange(0, 100) dates = np.arange(0, 100) win = 20 ### compare movavg calc to timeseries movmean: # ts mov_mean t = time_series(data=data, dates=dates) tMM = mov_mean(t, win) # finutils movingaverage (with leading zeros, like timeseries) fMA = np.array([0 if i < win-1 else np.mean(data[i-win+1:i+1]) for i in dates]) print 'moving average errors: %f' % (sum(tMM!=fMA)) ### compare exp movavg calc to timeseries expw movavg # ts expw tEXP = mov_average_expw(t, win) # finutils fEXP = fn.movingema(t, win, 0.0952381) print 'exp moving average errors: %f' % (sum(tEXP!=fEXP)) plot(tEXP-fEXP, '-g')
import numpy as np import matplotlib.pyplot as plt import scikits.timeseries as ts import scikits.timeseries.lib.plotlib as tpl from scikits.timeseries.lib.moving_funcs import mov_average_expw # generate some random data data = np.cumprod(1 + np.random.normal(0, 1, 300) / 100) series = ts.time_series(data, start_date=ts.Date(freq="M", year=1982, month=1)) fig = tpl.tsfigure() fsp = fig.add_tsplot(111) fsp.tsplot(series, "-", mov_average_expw(series, 40), "r--") plt.show()
from scikits.timeseries.lib.moving_funcs import mov_mean, mov_average_expw
import utilfuncs.finutils as fn
reload(fn)
from matplotlib.pyplot import plot
data = np.arange(0, 100)
dates = np.arange(0, 100)
win = 20
### compare movavg calc to timeseries movmean:
# ts mov_mean
t = time_series(data=data, dates=dates)
tMM = mov_mean(t, win)
# finutils movingaverage (with leading zeros, like timeseries)
fMA = np.array(
[0 if i < win - 1 else np.mean(data[i - win + 1:i + 1]) for i in dates])
print 'moving average errors: %f' % (sum(tMM != fMA))
### compare exp movavg calc to timeseries expw movavg
# ts expw
tEXP = mov_average_expw(t, win)
# finutils
fEXP = fn.movingema(t, win, 0.0952381)
print 'exp moving average errors: %f' % (sum(tEXP != fEXP))
plot(tEXP - fEXP, '-g')