def _predict_dir(self, x_dir):
sec_veer = []
for i in range(1, self.sectors + 1):
sec_veer.append(self.params[i]['average_veer'])
# Add additional entry for first sector
sec_veer.append(self.params[1]['average_veer'])
if self.direction_bin_array is None:
veer_bins = [
i * (360 / self.sectors) for i in range(0, self.sectors + 1)
]
else:
veer_bins = [
self.direction_bin_array[i] +
self.direction_bin_array[i + 1] / 2.0
for i in range(0,
len(self.direction_bin_array) - 1)
]
x = pd.concat([
x_dir.dropna().rename('dir'),
_binned_direction_series(
x_dir.dropna(), self.sectors,
direction_bin_array=veer_bins).rename('veer_bin')
],
axis=1,
join='inner')
x['sec_mid_pt'] = [veer_bins[i - 1] for i in x['veer_bin']]
x['ratio'] = (x['dir'] - x['sec_mid_pt']) / (360.0 / self.sectors)
x['sec_veer'] = [sec_veer[i - 1] for i in x['veer_bin']]
x['multiply_factor'] = [
sec_veer[i] - sec_veer[i - 1] for i in x['veer_bin']
]
x['adjustment'] = x['sec_veer'] + (x['ratio'] * x['multiply_factor'])
return (x['dir'] + x['adjustment']).sort_index().apply(
utils._range_0_to_360)
def _predict(self, x_spd, x_dir):
x = pd.concat([
x_spd.dropna().rename('spd'),
_binned_direction_series(x_dir.dropna(),
self.sectors,
direction_bin_array=self.
direction_bin_array).rename('ref_dir_bin')
],
axis=1,
join='inner')
prediction = pd.DataFrame()
first = True
for sector, data in x.groupby(['ref_dir_bin']):
if first is True:
first = False
prediction = self.speed_model[sector].sector_predict(
data['spd'])
else:
prediction = pd.concat([
prediction, self.speed_model[sector].sector_predict(
data['spd'])
],
axis=0)
return prediction.sort_index()
def __init__(self, ref_spd, ref_dir, target_spd, target_dir, averaging_prd, coverage_threshold=0.9, sectors=12,
direction_bin_array=None, lt_ref_speed=None, preprocess=True):
CorrelBase.__init__(self, ref_spd, target_spd, averaging_prd, coverage_threshold, ref_dir=ref_dir,
target_dir=target_dir, preprocess=preprocess)
self.sectors = sectors
self.direction_bin_array = direction_bin_array
if direction_bin_array is not None:
self.sectors = len(direction_bin_array)-1
if lt_ref_speed is None:
self.lt_ref_speed = momm(self.data['ref_spd'])
else:
self.lt_ref_speed = lt_ref_speed
self.cutoff = min(0.5 * self.lt_ref_speed, 4.0)
self.ref_veer_cutoff = self._get_veer_cutoff(self.data['ref_spd'])
self.target_veer_cutoff = self._get_veer_cutoff((self.data['target_spd']))
self._randomize_calm_periods()
self._get_overall_veer()
# for low ref_speed and high target_speed recalculate direction sector
self._adjust_low_reference_speed_dir()
self.ref_dir_bins = _binned_direction_series(self.data['ref_dir'], sectors,
direction_bin_array=self.direction_bin_array).rename('ref_dir_bin')
self.data = pd.concat([self.data, self.ref_dir_bins], axis=1, join='inner')
self.data = self.data.dropna()
self.params = 'not run yet'