def testRegressionWithRorSelectedNonalignedMultidimData(self):
"""Run a single experiment with multidimension data"""
data = self.createTimeSeriesData(10, 3, [np.random.randint(100, 1000) for unused in range(10)])
custom_vars = dict({"output_fncs": [lambda ds: np.mean(ds[1])], "data_mapping": DataNormalizer(0, 2)})
exp = RegressionExperiment()
val = exp.run(data, graph=False, disp=False, static_variables=custom_vars)
assert type(val) == np.float64, "Run did not return a number."
def testRegressionWithRorSelectedMultidimData(self):
"""Run a single experiment with multidimension data"""
data = self.createTimeSeriesData(150, 3, 50)
custom_vars = dict({"output_fncs": [lambda ds: np.mean(ds[1])]})
exp = RegressionExperiment()
val = exp.run(data, graph=False, disp=False, static_variables=custom_vars)
assert type(val) == np.float64, "Run did not return a number."
def testOneDimExperiment(self, repeats=0):
"""Test a one-dimensional regression experiment"""
exp = RegressionExperiment()
data_set = self.createTimeSeriesData(100, 200)
custom_vars = dict({"seg:validation_split": 0.3, "fg:generators": [gens.identity, gens.maximum, gens.minimum]})
indep_vars = dict({"seg:allowable_overlap": range(7, 9)})
val = exp.run(
data_set,
graph=False,
disp=False,
variables_under_test=indep_vars,
static_variables=custom_vars,
repeats=repeats,
)
assert type(val) == np.ndarray, "Run did not return a np.ndarray."
assert len(val) == len(indep_vars.values()[0]), "Run did not return a correct length array."
assert type(val[0]) == np.float64, "Run did not return a list of floats."
def testTwoDimExperiment(self, repeats=0):
"""Test a two-dimensional regression experiment"""
exp = RegressionExperiment()
data_set = self.createTimeSeriesData(100, 200)
custom_vars = dict({"seg:validation_split": 0.3, "fg:generators": [gens.identity, gens.maximum, gens.minimum]})
indep_vars = dict({"seg:allowable_overlap": range(1, 3), "seg:predictor_length": [15, 20]})
val = exp.run(
data_set,
graph=False,
disp=False,
variables_under_test=indep_vars,
static_variables=custom_vars,
repeats=repeats,
)
assert type(val) == np.ndarray, "Run did not return a np.ndarray."
assert np.shape(val) == (len(indep_vars.values()[0]), len(indep_vars.values()[1])) or np.shape(
np.transpose(val)
) == (len(indep_vars.values()[0]), len(indep_vars.values()[1])), "Run did not return a correct size array."
assert type(val[0][0]) == np.float64, "Run did not return an array of floats."
def testNoExperimentCustomVars(self, repeats=0):
"""Test a regression experiment with custom vars."""
exp = RegressionExperiment()
data_set = self.createTimeSeriesData(100, 200)
custom_vars = dict(
{
"output_fncs": [np.min],
"data_mapping": None,
"seg:predictor_length": 10,
"seg:predictee_length": 1,
"seg:allowable_overlap": 1,
"seg:validation_split": 0.3,
"fg:generators": [gens.identity, gens.maximum, gens.minimum],
"fg:threshold": 0.05,
"reg:constructor": SupportVectorRegressor,
"reg:training_params": dict({"kernel_type": 1, "degree": 2}),
}
)
val = exp.run(data_set, graph=False, disp=False, static_variables=custom_vars, repeats=repeats)
assert type(val) == np.float64, "Run did not return a number."
def __init__(self, arguments=sys.argv):
usage = 'usage: %prog [options] DATA_FILE'
parser = OptionParser(usage=usage)
# Input configuration
parser.add_option('-i','--input-file',dest='input_files',action='append',
type='string',metavar='FILE',
help='Configuration input file for a single run')
parser.add_option('-f','--format',dest='format',action='store',choices=self.readers.keys(),
metavar='FMT',help='Data File Format',default='csv')
parser.add_option('-o','--output-file',dest='output_file',action='store',type='string',
default='regresion_experiment.csv',help='Output File',metavar='FILE')
parser.add_option('-q','--quiet',dest='verbose',action='store_false',type='bool',
default=True,help='Quiet the run')
(options, datafile) = parser.parse_args(arguments)
data = self.readers[options['format']](datafile)
exp = RegressionExperiment()
if options['input_files'] == []:
print 'Running regression experiment using no input file'
print exp.run(data.quotes(), graph=options['verbose'], disp=options['verbose'])
else:
for ifn in options['input_files']:
print 'Running regression experiment using input file:', ifn
(static_config, variable_config) = deserializeConfiguration(ifn)
print exp.run(
data.quotes(),
graph=options['verbose'],
disp=options['verbose'],
static_variables=static_config,
variables_under_test=variable_config)
from minerva.regression import SupportVectorRegressor
datadir = '/home/jon/Code/Minerva/test_data/'
datafile = 'test_data_nasdaq_a'
# datafile = 'nasdaq_full_1990_to_2010'
testvars = dict({'seg:predictor_length': range(10,50,2)})
nontestvars = dict({
'fg:generators': [[gens.identity],
[gens.identity,gens.mean,gens.stdev],
[gens.identity],
[gens.identity],
[gens.identity,gens.maximum,gens.minimum]],
'data_mapping': DataNormalizer(ror_divisor_row=1, volume_row=4),
'output_fncs': [lambda v: np.mean(v[1]), lambda v: np.std(v[1])],
'reg:constructor': SupportVectorRegressor,
'reg:training_params': dict({'use_shrinking': '0'})
})
repeats = 10
if __name__ == '__main__':
data_file = QuoteReaderCsv(datadir + datafile)
exp = RegressionExperiment()
outdata = exp.run(data_file.quotes(),
graph=True,
disp=True,
variables_under_test=testvars,
static_variables=nontestvars,
repeats=repeats)
def testNoExperimentDefaultVars(self, repeats=0):
"""Test a regression experiment with default vars."""
exp = RegressionExperiment()
data_set = self.createTimeSeriesData(100, 200)
val = exp.run(data_set, graph=False, disp=False, repeats=repeats)
assert val * 0 == 0, "Run did not return a number."
def testRegressionWithNonAlignedData(self):
"""Run a single experiment with nonaligned data"""
data = self.createTimeSeriesData(10, [np.random.randint(100, 1000) for unused in range(10)])
exp = RegressionExperiment()
val = exp.run(data, graph=False, disp=False)
assert type(val) == np.float64, "Run did not return a number."