def run(self):
blendranges = []
submodel_blend_names = []
submodel_predictions = []
datakey = self.chooseDatacomboBox.currentText()
refcol = ('comp', self.optimizeSubRangesComboBox.currentText())
# start with the low model
blendranges.append(
[-9999, float(self.lowPredictionMaxSpinBox.value())])
submodel_blend_names.append(self.lowPredictionComboBox.currentText())
submodel_predictions.append(
self.data[datakey].df[('predict',
self.lowPredictionComboBox.currentText())])
# append the intermediate submodels
for i in range(0, self.index):
temp_vals = self.subwidgets[i].getValues()
blendranges.append(temp_vals[1])
submodel_blend_names.append(temp_vals[0])
submodel_predictions.append(self.data[datakey].df[('predict',
temp_vals[0])])
# append the high model
blendranges.append(
[float(self.highPredictionMinSpinBox.value()), 9999])
submodel_blend_names.append(self.highPredictionComboBox.currentText())
submodel_predictions.append(
self.data[datakey].df[('predict',
self.highPredictionComboBox.currentText())])
# append the reference model as a catch-all
blendranges.append([-9999, 9999])
submodel_blend_names.append(
self.referencePredictionComboBox.currentText())
submodel_predictions.append(self.data[datakey].df[(
'predict', self.referencePredictionComboBox.currentText())])
if self.optimizeSubRangesCheckBox.isChecked():
truevals = self.data[datakey].df[refcol]
else:
truevals = None
sm_obj = sm.sm(blendranges)
# optimize blending if reference data is provided
if truevals is not None:
predictions_blended = sm_obj.do_blend(submodel_predictions,
truevals=truevals)
else:
# otherwise just blend the predictions together
predictions_blended = sm_obj.do_blend(submodel_predictions)
# save the blended predictions
self.data[datakey].df[('predict', 'Blended-Predict ' +
str(sm_obj.blendranges))] = predictions_blended
def setup(self):
blendranges = []
datakey = self.chooseDataComboBox.currentText()
# start with the low model
blendranges.append(
[-9999, float(self.lowPredictionMaxSpinBox.value())])
# append the intermediate submodels
for i in range(0, self.index_spin.value()):
temp_vals = self.subwidgets[i].getValues()
blendranges.append(temp_vals[1])
# append the high model
blendranges.append(
[float(self.highPredictionMinSpinBox.value()), 9999])
sm_obj = sm.sm(blendranges)
# save the blended predictions
try:
self.data[datakey].df[('predict', 'Blended-Predict ' +
str(sm_obj.blendranges))] = 99999
except:
pass
def run(self):
blendranges = []
submodel_names = []
x_ref = []
x = []
submodels = []
# TODO there are some inefficiencies with this code:103:111
# For example you are running through the UI to collect the data
# And then running through the data to collect more data
# You're doing this with seperate for loops. this could be done
# with just one for loop. Example lines 103 and 111.
self.submodel_gui_info = [[
self.lowModelComboBox.currentText(),
[-9999, int(self.lowModelMaxSpinBox.value())]
]]
for i in range(1, self.index):
self.submodel_gui_info.append(self.subwidgets[i].getValues())
self.submodel_gui_info.append([
self.highModelComboBox.currentText(),
[int(self.highModelMinSpinBox.value()), 9999]
])
self.submodel_gui_info.append(
[self.referenceModelComboBox.currentText(), [-9999, 9999]])
datakey = self.chooseDataComboBox.currentText()
for sub_gui in self.submodel_gui_info:
min_temp = sub_gui[1][0]
max_temp = sub_gui[1][1]
blendranges.append([min_temp, max_temp])
submodel_names.append(sub_gui[0])
if self.optimizeSubmodelRangesCheckBox.isChecked():
trueval_data = self.optimizeSubmodelRangesComboBox.currentText()
else:
trueval_data = None
# Check if reference data name has been provided
# if so, get reference data values
if trueval_data is not None:
truevals = self.data[trueval_data].df[self.model_yvars[
submodel_names[0]]]
else:
truevals = None
# step through the submodel names and get the actual models and the x data
for i in submodel_names:
x.append(self.data[datakey].df[self.model_xvars[i]])
submodels.append(self.models[i])
if trueval_data is not None:
x_ref.append(self.data[trueval_data].df[self.model_xvars[i]])
# create the submodel object
sm_obj = sm.sm(blendranges, submodels)
# optimize blending if reference data is provided (otherwise, modelranges will be used as blending ranges)
if truevals is not None:
ref_predictions = sm_obj.predict(x_ref)
ref_predictions_blended = sm_obj.do_blend(ref_predictions,
truevals=truevals)
# get predictions for each submodel separately
predictions = sm_obj.predict(x)
# blend the predictions together
predictions_blended = sm_obj.do_blend(predictions)
# save the individual and blended predictions
for i, j in enumerate(predictions):
self.data[datakey].df[('predict',
submodel_names[i] + '-Predict')] = j
self.data[datakey].df[('predict',
'Blended-Predict')] = predictions_blended