def trainModel(self):
if not utils.checkFactors(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Factors rasters is not set. Please specify them and try again")
)
return
matrix = self.__checkMatrix()
if len(matrix) == 0:
QMessageBox.warning(self.plugin,
self.tr("Incorrect matrix"),
self.tr("Please fill the matrix with values")
)
return
self.settings.setValue("ui/MCE/initialClass", self.spnInitialClass.value())
self.settings.setValue("ui/MCE/finalClass", self.spnFinalClass.value())
areaAnalyst = AreaAnalyst(self.inputs["initial"], second=None)
self.plugin.logMessage(self.tr("Init MCE model"))
model = MCE(self.inputs["factors"].values(),
matrix,
self.spnInitialClass.value(),
self.spnFinalClass.value(),
areaAnalyst
)
self.inputs["model"] = model
self.plugin.logMessage(self.tr("MCE model trained"))
weights = model.getWeights()
for i, w in enumerate(weights):
item = QTableWidgetItem(unicode(w))
self.tblWeights.setItem(0, i, item)
self.tblWeights.resizeRowsToContents()
self.tblWeights.resizeColumnsToContents()
# Check consistency of the matrix
c = model.getConsistency()
if c < 0.1:
QMessageBox.warning(self.plugin,
self.tr("Consistent matrix"),
self.tr("Matrix filled correctly. Consistency value is: %f. The model can be used.") % (c)
)
else:
QMessageBox.warning(self.plugin,
self.tr("Inconsistent matrix"),
self.tr("Please adjust matrix before starting simulation. Consistency value is: %f") % (c)
)
def trainModel(self):
if not utils.checkInputRasters(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Initial or final raster is not set. Please specify input data and try again")
)
return
if not utils.checkFactors(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Factors rasters is not set. Please specify them and try again")
)
return
myBins = self.__getBins()
self.plugin.logMessage(self.tr("Init WoE model"))
try:
model = WoeManager(self.inputs["factors"].values(), self.plugin.analyst, bins=myBins)
except WoeManagerError as err:
QMessageBox.warning(self.plugin,
self.tr("Initialization error"),
err.msg
)
return
self.inputs["model"] = model
if not model.checkBins():
QMessageBox.warning(self.plugin,
self.tr("Wrong ranges"),
self.tr("Ranges are not correctly specified. Please specify them and try again (use space as separator)")
)
return
model.moveToThread(self.plugin.workThread)
self.plugin.workThread.started.connect(model.train)
model.updateProgress.connect(self.plugin.showProgress)
model.rangeChanged.connect(self.plugin.setProgressRange)
model.errorReport.connect(self.plugin.logErrorReport)
model.processFinished.connect(self.__trainFinished)
model.processFinished.connect(self.plugin.workThread.quit)
self.plugin.workThread.start()
def startFitModel(self):
if not utils.checkInputRasters(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Initial or final raster is not set. Please specify input data and try again")
)
return
if not utils.checkFactors(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Factors rasters is not set. Please specify them and try again")
)
return
if not utils.checkChangeMap(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Change map raster is not set. Please create it try again")
)
return
self.settings.setValue("ui/LR/neighborhood", self.spnNeighbourhood.value())
self.plugin.logMessage(self.tr("Init LR model"))
model = LR(ns=self.spnNeighbourhood.value())
self.inputs["model"] = model
model.setMaxIter(self.spnMaxIterations.value())
model.setState(self.inputs["initial"])
model.setFactors(self.inputs["factors"].values())
model.setOutput(self.inputs["changeMap"])
model.setMode(self.inputs["samplingMode"],)
model.setSamples(self.plugin.spnSamplesCount.value())
self.plugin.logMessage(self.tr("Set training data"))
model.moveToThread(self.plugin.workThread)
self.plugin.workThread.started.connect(model.startTrain)
self.plugin.setProgressRange("Train LR model", 0)
model.finished.connect(self.__trainFinished)
model.errorReport.connect(self.plugin.logErrorReport)
model.finished.connect(self.plugin.workThread.quit)
self.plugin.workThread.start()
def manageGui(self):
if not utils.checkFactors(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Factors rasters is not set. Please specify them and try again")
)
return
if not utils.checkChangeMap(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Change map raster is not set. Please create it try again")
)
return
self.tblReclass.clearContents()
self.delegate = spinboxdelegate.SpinBoxDelegate(self.tblReclass.model(), minRange=2, maxRange=dataprovider.MAX_CATEGORIES)
row = 0
for k, v in self.inputs["factors"].iteritems():
v.denormalize() # Denormalize the factor's bands if they are normalized
for b in xrange(1, v.getBandsCount()+1):
if v.isCountinues(b):
self.tblReclass.insertRow(row)
if v.getBandsCount()>1:
name = u"%s (band %s)" % (utils.getLayerById(k).name(), unicode(b))
else:
name = u"%s" % (utils.getLayerById(k).name(), )
stat = v.getBandStat(b)
for n, item_data in enumerate([name, (u"%f" % (stat["min"], )), (u"%f" % (stat["max"])), u"" , u"" ]):
item = QTableWidgetItem(item_data)
if n < 3:
item.setFlags(item.flags() ^ Qt.ItemIsEditable)
self.tblReclass.setItem(row, n, item)
row += 1
rowCount = row
self.tblReclass.setItemDelegateForColumn(3, self.delegate)
for row in range(rowCount):
# Set 2 bins as default value
self.tblReclass.setItem(row, 3, QTableWidgetItem(u'2'))
self.tblReclass.resizeRowsToContents()
self.tblReclass.resizeColumnsToContents()
def manageGui(self):
if not utils.checkInputRasters(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Initial or final raster is not set. Please specify input data and try again")
)
return
if not utils.checkFactors(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Factors rasters is not set. Please specify them and try again")
)
return
self.spnInitialClass.setValue(int(self.settings.value("ui/MCE/initialClass", 0)))
self.spnFinalClass.setValue(int(self.settings.value("ui/MCE/finalClass", 0)))
gradations = self.inputs["initial"].getBandGradation(1)
self.spnInitialClass.setRange(min(gradations), max(gradations))
gradations = self.inputs["final"].getBandGradation(1)
self.spnFinalClass.setRange(min(gradations), max(gradations))
self.__prepareTable()
def trainNetwork(self):
if not utils.checkInputRasters(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Initial or final raster is not set. Please specify input data and try again")
)
return
if not utils.checkFactors(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Factors rasters is not set. Please specify them and try again")
)
return
if not utils.checkChangeMap(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Change map raster is not set. Please create it try again")
)
return
if self.leTopology.text() == "":
QMessageBox.warning(self.plugin,
self.tr("Wrong network topology"),
self.tr("Network topology is undefined. Please define it and try again")
)
return
self.settings.setValue("ui/ANN/neighborhood", self.spnNeigbourhood.value())
self.settings.setValue("ui/ANN/learningRate", self.spnLearnRate.value())
self.settings.setValue("ui/ANN/maxIterations", self.spnMaxIterations.value())
self.settings.setValue("ui/ANN/topology", self.leTopology.text())
self.settings.setValue("ui/ANN/momentum", self.spnMomentum.value())
self.btnStop.setEnabled(True)
self.btnTrainNetwork.setEnabled(False)
self.plugin.logMessage(self.tr("Init ANN model"))
model = MlpManager(ns=self.spnNeigbourhood.value())
self.inputs["model"] = model
model.createMlp(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
[int(n) for n in self.leTopology.text().split(" ")]
)
self.plugin.logMessage(self.tr("Set training data"))
model.setTrainingData(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
mode=self.inputs["samplingMode"],
samples=self.plugin.spnSamplesCount.value()
)
model.setEpochs(self.spnMaxIterations.value())
model.setValPercent(20)
model.setLRate(self.spnLearnRate.value())
model.setMomentum(self.spnMomentum.value())
model.setContinueTrain()
self.axes.cla()
self.axes.grid(True)
self.dataTrain = []
self.dataVal = []
self.plotTrain = self.axes.plot(self.dataTrain,
linewidth=1,
color="green", marker='o'
)[0]
self.plotVal = self.axes.plot(self.dataVal,
linewidth=1,
color="red",
)[0]
leg = self.axes.legend(('Train', 'Validation'), 'upper right', shadow=False)
for t in leg.get_texts():
t.set_fontsize('small')
model.moveToThread(self.plugin.workThread)
self.plugin.workThread.started.connect(model.startTrain)
self.btnStop.clicked.connect(model.stopTrain)
model.updateGraph.connect(self.__updateGraph)
model.updateDeltaRMS.connect(self.__updateRMS)
model.updateMinValErr.connect(self.__updateValidationError)
model.updateKappa.connect(self.__updateKappa)
model.processInterrupted.connect(self.__trainInterrupted)
model.rangeChanged.connect(self.plugin.setProgressRange)
model.updateProgress.connect(self.plugin.showProgress)
model.errorReport.connect(self.plugin.logErrorReport)
model.processFinished.connect(self.__trainFinished)
model.processFinished.connect(self.plugin.workThread.quit)
self.plugin.logMessage(self.tr("Start trainig ANN model"))
self.plugin.workThread.start()
def trainNetwork(self):
if not utils.checkInputRasters(self.inputs):
QMessageBox.warning(
self.plugin, self.tr("Missed input data"),
self.
tr("Initial or final raster is not set. Please specify input data and try again"
))
return
if not utils.checkFactors(self.inputs):
QMessageBox.warning(
self.plugin, self.tr("Missed input data"),
self.
tr("Factors rasters is not set. Please specify them and try again"
))
return
if not utils.checkChangeMap(self.inputs):
QMessageBox.warning(
self.plugin, self.tr("Missed input data"),
self.
tr("Change map raster is not set. Please create it try again"))
return
if self.leTopology.text() == "":
QMessageBox.warning(
self.plugin, self.tr("Wrong network topology"),
self.
tr("Network topology is undefined. Please define it and try again"
))
return
self.settings.setValue("ui/ANN/neighborhood",
self.spnNeigbourhood.value())
self.settings.setValue("ui/ANN/learningRate",
self.spnLearnRate.value())
self.settings.setValue("ui/ANN/maxIterations",
self.spnMaxIterations.value())
self.settings.setValue("ui/ANN/topology", self.leTopology.text())
self.settings.setValue("ui/ANN/momentum", self.spnMomentum.value())
self.btnStop.setEnabled(True)
self.btnTrainNetwork.setEnabled(False)
self.plugin.logMessage(self.tr("Init ANN model"))
model = MlpManager(ns=self.spnNeigbourhood.value())
self.inputs["model"] = model
model.createMlp(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
[int(n) for n in self.leTopology.text().split(" ")])
self.plugin.logMessage(self.tr("Set training data"))
model.setTrainingData(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
mode=self.inputs["samplingMode"],
samples=self.plugin.spnSamplesCount.value())
model.setEpochs(self.spnMaxIterations.value())
model.setValPercent(20)
model.setLRate(self.spnLearnRate.value())
model.setMomentum(self.spnMomentum.value())
model.setContinueTrain()
self.axes.cla()
self.axes.grid(True)
self.dataTrain = []
self.dataVal = []
self.plotTrain = self.axes.plot(self.dataTrain,
linewidth=1,
color="green",
marker='o')[0]
self.plotVal = self.axes.plot(
self.dataVal,
linewidth=1,
color="red",
)[0]
leg = self.axes.legend(('Train', 'Validation'),
'upper right',
shadow=False)
for t in leg.get_texts():
t.set_fontsize('small')
model.moveToThread(self.plugin.workThread)
self.plugin.workThread.started.connect(model.startTrain)
self.btnStop.clicked.connect(model.stopTrain)
model.updateGraph.connect(self.__updateGraph)
model.updateDeltaRMS.connect(self.__updateRMS)
model.updateMinValErr.connect(self.__updateValidationError)
model.updateKappa.connect(self.__updateKappa)
model.processInterrupted.connect(self.__trainInterrupted)
model.rangeChanged.connect(self.plugin.setProgressRange)
model.updateProgress.connect(self.plugin.showProgress)
model.errorReport.connect(self.plugin.logErrorReport)
model.processFinished.connect(self.__trainFinished)
model.processFinished.connect(self.plugin.workThread.quit)
self.plugin.logMessage(self.tr("Start trainig ANN model"))
self.plugin.workThread.start()