def testSecurityMovingCorrelationNested(self):
window = 120
x = SecurityLatestValueHolder('close')
y = SecurityLatestValueHolder('open')
z = x > y
mc = SecurityMovingCorrelation(window, x, z)
for i in range(len(self.aapl['close'])):
data = dict(aapl=dict(close=self.aapl['close'][i],
open=self.aapl['open'][i]),
ibm=dict(close=self.ibm['close'][i],
open=self.ibm['open'][i]))
mc.push(data)
if i < window:
start = 0
else:
start = i + 1 - window
if i < 1:
continue
value = mc.value
for i, name in enumerate(value.index()):
if i >= window:
xs = self.dataSet[name]['close'][start:(i + 1)]
ys = self.dataSet[name]['open'][start:(i + 1)]
zs = np.where(xs > ys, 1., 0.)
expected = np.corrcoef(xs, zs)[0, 1]
calculated = value[name]
self.assertAlmostEqual(
expected, calculated, 12, 'at index {0}\n'
'expected: {1:.12f}\n'
'calculated: {2:.12f}'.format(i, expected, calculated))
def testSecurityMovingCorrelation(self):
window = 100
mr = SecurityMovingCorrelation(window, ('y', 'x'))
for i in range(len(self.aapl['close'])):
data = {
'aapl': {
'y': self.aapl['close'][i],
'x': self.aapl['open'][i]
},
'ibm': {
'y': self.ibm['close'][i],
'x': self.ibm['open'][i]
}
}
mr.push(data)
if i >= window - 1:
for name in mr.value.index():
series_x = getattr(self,
name)['open'][i - window + 1:i + 1]
series_y = getattr(self,
name)['close'][i - window + 1:i + 1]
expected_res = np.corrcoef(series_x, series_y)[0, 1]
self.assertAlmostEqual(mr.value[name], expected_res, msg= \
"at index {0} and symbol {1}\n"
"expected: {2}\n"
"calculated: {3}".format(i, name, expected_res, mr.value[name]))
def testSecurityMovingCorrelationPickle(self):
ma = SecurityMovingCorrelation(10, ['y', 'x'])
data = dict(aapl=dict(x=1., y=2), ibm=dict(x=2., y=3))
data2 = dict(aapl=dict(x=2., y=3), ibm=dict(x=3., y=4))
ma.push(data)
ma.push(data2)
with tempfile.NamedTemporaryFile('w+b', delete=False) as f:
pickle.dump(ma, f)
with open(f.name, 'rb') as f2:
pickled = pickle.load(f2)
os.unlink(f.name)
for i, v in enumerate(np.random.rand(20)):
data['aapl']['x'] = v
data['aapl']['y'] = v + 1
data['ibm']['x'] = v + 1.
data['ibm']['y'] = v + 2.
ma.push(data)
pickled.push(data)
if i >= 10:
for name in ma.value.index():
self.assertAlmostEqual(ma.value[name], pickled.value[name])
def testSecurityMovingCorrelationDeepcopy(self):
ma = SecurityMovingCorrelation(10, ['y', 'x'])
data = dict(aapl=dict(x=1., y=2), ibm=dict(x=2., y=3))
data2 = dict(aapl=dict(x=2., y=3), ibm=dict(x=3., y=4))
ma.push(data)
ma.push(data2)
copied = copy.deepcopy(ma)
for i, v in enumerate(np.random.rand(20)):
data['aapl']['x'] = v
data['aapl']['y'] = v + 1
data['ibm']['x'] = v + 1.
data['ibm']['y'] = v + 2.
ma.push(data)
copied.push(data)
if i >= 10:
for name in ma.value.index():
self.assertAlmostEqual(ma.value[name], copied.value[name])
def CORR(window, dependency=('x', 'y')):
return SecurityMovingCorrelation(window, dependency)
def MCORR(window, x='x', y='y'):
return SecurityMovingCorrelation(window, x, y)
