def test_binarization(self):
t = binaryzation(self.X, [0, 2])
answer1 = np.array([[
False,
True,
False,
], [
False,
True,
False,
], [True, False, True]])
answer2 = np.array([[
False,
True,
False,
], [
False,
True,
False,
], [True, False, True]])
assert_array_equal(t, answer1)
assert_array_equal(t, answer2)
mask = [[False, False, False], [False, False, False],
[True, False, False]]
data = [[False, True, False], [False, True, False],
[False, False, True]]
assert_array_equal(binaryzation(np.array(data), [True]),
np.ma.array(data=data, mask=mask))
data = np.ma.array(data=data, mask=mask)
assert_array_equal(binaryzation(data, [True]),
np.ma.array(data=data, mask=mask))
def _predict(self, state, calcTransitions=False):
'''
Predict the changes.
'''
try:
geodata = state.getGeodata()
rows, cols = geodata['ySize'], geodata['xSize']
self.transitionPotentials = None # Reset tr.potentials if they exist
# Get locations where self.initStateNum is occurs
band = state.getBand(1)
initStateMask = binaryzation(band, [self.initStateNum])
mask = band.mask
# Calculate summary map of factors weights
# Transition potentials:
# current implementation: potential and confidence are equal (two-class implementation)
# Confidence:
# confidence is summary map of factors scaled to 0-100, if current state = self.initState
# confidence is 0, if current state != self.initState
# Prediction:
# predicted value is a constant = areaAnalyst.encode(initStateNum, finalStateNum), if current state = self.initState
# predicted value is the transition code current_state -> current_state, if current state != self.initState
confidence = np.zeros((rows,cols), dtype=np.uint8)
weights = self.getWeights()
weightNum = 0 # Number of processed weights
for f in self.factors:
if not f.geoDataMatch(state):
raise MCEError('Geometries of the state and factor rasters are different!')
f.normalize(mode = 'maxmin')
for i in xrange(f.getBandsCount()):
band = f.getBand(i+1)
confidence = confidence + (band*weights[weightNum]*100).astype(np.uint8)
mask = np.ma.mask_or(mask, band.mask)
weightNum = weightNum + 1
confidence = confidence*initStateMask
prediction = np.copy(state.getBand(1))
for code in self.areaAnalyst.categories:
if code != self.initStateNum:
prediction[prediction==code] = self.areaAnalyst.encode(code, code)
else:
prediction[prediction==code] = self.areaAnalyst.encode(self.initStateNum, self.finalStateNum)
predicted_band = np.ma.array(data=prediction, mask=mask, dtype=np.uint8)
self.prediction = Raster()
self.prediction.create([predicted_band], geodata)
confidence_band = np.ma.array(data=confidence, mask=mask, dtype=np.uint8)
self.confidence = Raster()
self.confidence.create([confidence_band], geodata)
code = self.areaAnalyst.encode(self.initStateNum, self.finalStateNum)
self.transitionPotentials = {code: self.confidence}
except MemoryError:
self.errorReport.emit(self.tr("The system out of memory during MCE prediction"))
raise
except:
self.errorReport.emit(self.tr("An unknown error occurs during MCE prediction"))
raise
def woe(factor, sites, unit_cell=1):
'''Weight of evidence method (multiclass form).
@param factor Multiclass pattern array used for prediction of point objects (sites).
@param sites Array layer consisting of the locations at which the point objects are known to occur.
@param unit_cell Method parameter, pixelsize of resampled rasters.
@return masked array Array of total weights of each factor.
'''
# Get list of categories from the factor raster
categories = get_gradations(factor.compressed())
# Try to binarize sites:
sCategories = get_gradations(sites.compressed())
if len(sCategories) != 2:
raise WoeError('Site raster must be binary!')
sites = binaryzation(sites, [sCategories[1]])
# List of the weights of evidence:
# weights[0] is (wPlus, wMinus) for the first category, weights[1] is (wPlus, wMinus) for the second category, ...
weights = []
if len(categories) >= 2:
for cat in categories:
fct = binaryzation(factor, [cat])
weights.append(_binary_woe(fct, sites, unit_cell))
else:
raise WoeError('Wrong count of categories in the factor raster!')
wTotalMin = sum([w[1] for w in weights])
# List of total weights of evidence of the categories:
# wMap[0] is the total weight of the first category, wMap[1] is the total weight of the second category, ...
wMap = [w[0] + wTotalMin - w[1] for w in weights]
# If len(categories) = 2, then [w[0] + wTotalMin - w[1] for w in weights] increases the answer.
# In this case:
if len(categories) == 2:
wMap = [w/2 for w in wMap]
resultMap =np.zeros(ma.shape(factor))
for i,cat in enumerate(categories):
resultMap[factor==cat] = wMap[i]
resultMap = ma.array(data=resultMap, mask=factor.mask)
result = {'map': resultMap, 'categories': categories, 'weights': wMap}
return result
def woe(factor, sites, unit_cell=1):
'''Weight of evidence method (multiclass form).
@param factor Multiclass pattern array used for prediction of point objects (sites).
@param sites Array layer consisting of the locations at which the point objects are known to occur.
@param unit_cell Method parameter, pixelsize of resampled rasters.
@return masked array Array of total weights of each factor.
'''
# Get list of categories from the factor raster
categories = get_gradations(factor.compressed())
# Try to binarize sites:
sCategories = get_gradations(sites.compressed())
if len(sCategories) != 2:
raise WoeError('Site raster must be binary!')
sites = binaryzation(sites, [sCategories[1]])
# List of the weights of evidence:
# weights[0] is (wPlus, wMinus) for the first category, weights[1] is (wPlus, wMinus) for the second category, ...
weights = []
if len(categories) >= 2:
for cat in categories:
fct = binaryzation(factor, [cat])
weights.append(_binary_woe(fct, sites, unit_cell))
else:
raise WoeError('Wrong count of categories in the factor raster!')
wTotalMin = sum([w[1] for w in weights])
# List of total weights of evidence of the categories:
# wMap[0] is the total weight of the first category, wMap[1] is the total weight of the second category, ...
wMap = [w[0] + wTotalMin - w[1] for w in weights]
# If len(categories) = 2, then [w[0] + wTotalMin - w[1] for w in weights] increases the answer.
# In this case:
if len(categories) == 2:
wMap = [w / 2 for w in wMap]
resultMap = np.zeros(ma.shape(factor))
for i, cat in enumerate(categories):
resultMap[factor == cat] = wMap[i]
resultMap = ma.array(data=resultMap, mask=factor.mask)
result = {'map': resultMap, 'categories': categories, 'weights': wMap}
return result
def __init__(self, referenceMap, simulatedMap):
"""
@param referenceMap Reference raster
@param simulatedMap Simulated raster
"""
QObject.__init__(self)
if referenceMap.getBandsCount() + simulatedMap.getBandsCount() != 2:
raise EBError(
'The reference and simulated rasters must be 1-band rasters!')
if not referenceMap.geoDataMatch(simulatedMap):
raise EBError(
'Geometries of the reference and simulated rasters are different!'
)
self.categories = referenceMap.getBandGradation(1)
for s in simulatedMap.getBandGradation(1):
if not s in self.categories:
raise EBError(
'Categories in the reference and simulated rasters are different!'
)
R = referenceMap.getBand(1)
S = simulatedMap.getBand(1)
self.shape = R.shape
R, S = masks_identity(R, S, dtype=np.uint8)
# Array for weight
self.W = np.ones(self.shape)
self.W = self.W - np.ma.getmask(R)
R = np.ma.filled(R, 0)
S = np.ma.filled(S, 0)
# Proportion of category j in pixel n at the beginning resolution of the reference map
self.Rj = {}
for j in self.categories:
self.Rj[j] = 1.0 * binaryzation(R, [j])
# Proportion of category j in pixel n at the beginning resolution of the simulated map
self.Sj = {}
for j in self.categories:
self.Sj[j] = 1.0 * binaryzation(S, [j])
def train(self):
"""
Train the model
"""
self.transitionPotentials = {}
try:
iterCount = len(self.codes) * len(self.factors)
self.rangeChanged.emit(self.tr("Training WoE... %p%"), iterCount)
changeMap = self.changeMap.getBand(1)
for code in self.codes:
sites = binaryzation(changeMap, [code])
# Reclass factors (continuous factor -> ordinal factor)
wMap = np.ma.zeros(changeMap.shape) # The map of summary weight of the all factors
self.weights[code] = {} # Dictionary for storing wheights of every raster's band
for k in xrange(len(self.factors)):
fact = self.factors[k]
self.weights[code][k] = {} # Weights of the factor
factorW = self.weights[code][k]
if self.bins: # Get bins of the factor
bin = self.bins[k]
if (bin != None) and fact.getBandsCount() != len(bin):
raise WoeManagerError("Count of bins list for multiband factor is't equal to band count!")
else:
bin = None
for i in range(1, fact.getBandsCount() + 1):
band = fact.getBand(i)
if bin and bin[i - 1]: #
band = reclass(band, bin[i - 1])
band, sites = masks_identity(band, sites, dtype=np.uint8) # Combine masks of the rasters
woeRes = woe(
band, sites, self.unit_cell
) # WoE for the 'code' (initState->finalState) transition and current 'factor'.
weights = woeRes["map"]
wMap = wMap + weights
factorW[i] = woeRes["weights"]
self.updateProgress.emit()
# Reclassification finished => set WoE coefficients
self.woe[code] = wMap # WoE for all factors and the transition code.
# Potentials are WoE map rescaled to 0--100 percents
band = (sigmoid(wMap) * 100).astype(np.uint8)
p = Raster()
p.create([band], self.geodata)
self.transitionPotentials[code] = p
gc.collect()
except MemoryError:
self.errorReport.emit("The system out of memory during WoE trainig")
raise
except:
self.errorReport.emit(self.tr("An unknown error occurs during WoE trainig"))
raise
finally:
self.processFinished.emit()
def __init__(self, factors, areaAnalyst, unit_cell=1, bins = None):
'''
@param factors List of the pattern rasters used for prediction of point objects (sites).
@param areaAnalyst AreaAnalyst that contains map of the changes, encodes and decodes class numbers.
@param unit_cell Method parameter, pixelsize of resampled rasters.
@param bins Dictionary of bins. Bins are binning boundaries that used for reduce count of classes.
For example if factors = [f0, f1], then bins could be (for example) {0:[bins for f0], 1:[bins for f1]} = {0:[[10, 100, 250]],1:[[0.2, 1, 1.5, 4]]}.
List of list used because a factor can be a multiband raster, we need get a list of bins for every band. For example:
factors = [f0, 2-band-factor], bins= {0: [[10, 100, 250]], 1:[[0.2, 1, 1.5, 4], [3, 4, 7]] }
'''
self.factors = factors
self.analyst = areaAnalyst
self.changeMap = areaAnalyst.getChangeMap()
self.prediction = None
self.confidence = None
if (bins != None) and (len(factors) != len(bins.keys())):
raise WoeManagerError('Lengths of bins and factors are different!')
for r in self.factors:
if not self.changeMap.geoDataMatch(r):
raise WoeManagerError('Geometries of the input rasters are different!')
if self.changeMap.getBandsCount() != 1:
raise WoeManagerError('Change map must have one band!')
# Get list of codes from the changeMap raster
classes = self.changeMap.getBandStat(1)['gradation']
cMap = self.changeMap.getBand(1)
self.codes = [int(c) for c in classes] # Codes of transitions initState->finalState (see AreaAnalyst.encode)
self.woe = {}
for code in self.codes:
sites = binaryzation(cMap, [code])
# TODO: reclass factors (continuous factor -> ordinal factor)
wMap = np.ma.zeros(cMap.shape)
for k in xrange(len(factors)):
fact = factors[k]
if bins: # Get bins of the factor
bin = bins[k]
if (bin != None) and fact.getBandsCount() != len(bin):
raise WoeManagerError("Count of bins list for multiband factor is't equal to band count!")
else: bin = None
for i in range(1, fact.getBandsCount()+1):
band = fact.getBand(i)
if bin:
band = reclass(band, bin[i-1])
band, sites = masks_identity(band, sites) # Combine masks of the rasters
weights = woe(band, sites, unit_cell) # WoE for the 'code' (initState->finalState) transition and current 'factor'.
wMap = wMap + weights
self.woe[code]=wMap # WoE for all factors and the transition.
def woe(factor, sites, unit_cell=1):
'''Weight of evidence method (multiclass form).
@param factor Multiclass pattern array used for prediction of point objects (sites).
@param sites Array layer consisting of the locations at which the point objects are known to occur.
@param unit_cell Method parameter, pixelsize of resampled rasters.
@return [wMap1, wMap2, ...] Total weights of each factor.
'''
result =np.zeros(ma.shape(factor))
# Get list of classes from the factor raster
classes = get_gradations(factor.compressed())
# Try to binarize sites:
sClasses = get_gradations(sites.compressed())
if len(sClasses) != 2:
raise WoeError('Site raster must be binary!')
sites = binaryzation(sites, [sClasses[1]])
weights = [] # list of the weights of evidence
if len(classes) >= 2:
for cl in classes:
fct = binaryzation(factor, [cl])
weights.append(_binary_woe(fct, sites, unit_cell))
else:
raise WoeError('Wrong count of classes in the factor raster!')
wTotalMin = sum([w[1] for w in weights])
wMap = [w[0] + wTotalMin - w[1] for w in weights]
# If len(classes) = 2, then [w[0] + wTotalMin - w[1] for w in weights] increases the answer.
# In this case:
if len(classes) == 2:
wMap = [w/2 for w in wMap]
for i,cl in enumerate(classes):
result[factor==cl] = wMap[i]
result = ma.array(data=result, mask=factor.mask)
return result
def train(self):
'''
Train the model
'''
self.transitionPotentials = {}
try:
iterCount = len(self.codes)*len(self.factors)
self.rangeChanged.emit(self.tr("Training WoE... %p%"), iterCount)
changeMap = self.changeMap.getBand(1)
for code in self.codes:
sites = binaryzation(changeMap, [code])
# Reclass factors (continuous factor -> ordinal factor)
wMap = np.ma.zeros(changeMap.shape) # The map of summary weight of the all factors
self.weights[code] = {} # Dictionary for storing wheights of every raster's band
for k in xrange(len(self.factors)):
fact = self.factors[k]
self.weights[code][k] = {} # Weights of the factor
factorW = self.weights[code][k]
if self.bins: # Get bins of the factor
bin = self.bins[k]
if (bin != None) and fact.getBandsCount() != len(bin):
raise WoeManagerError("Count of bins list for multiband factor is't equal to band count!")
else: bin = None
for i in range(1, fact.getBandsCount()+1):
band = fact.getBand(i)
if bin and bin[i-1]: #
band = reclass(band, bin[i-1])
band, sites = masks_identity(band, sites, dtype=np.uint8) # Combine masks of the rasters
woeRes = woe(band, sites, self.unit_cell) # WoE for the 'code' (initState->finalState) transition and current 'factor'.
weights = woeRes['map']
wMap = wMap + weights
factorW[i] = woeRes['weights']
self.updateProgress.emit()
# Reclassification finished => set WoE coefficients
self.woe[code]=wMap # WoE for all factors and the transition code.
# Potentials are WoE map rescaled to 0--100 percents
band = (sigmoid(wMap)*100).astype(np.uint8)
p = Raster()
p.create([band], self.geodata)
self.transitionPotentials[code] = p
gc.collect()
except MemoryError:
self.errorReport.emit('The system out of memory during WoE trainig')
raise
except:
self.errorReport.emit(self.tr("An unknown error occurs during WoE trainig"))
raise
finally:
self.processFinished.emit()
def __init__ (self, referenceMap, simulatedMap):
"""
@param referenceMap Reference raster
@param simulatedMap Simulated raster
"""
QObject.__init__(self)
if referenceMap.getBandsCount() + simulatedMap.getBandsCount() !=2:
raise EBError('The reference and simulated rasters must be 1-band rasters!')
if not referenceMap.geoDataMatch(simulatedMap):
raise EBError('Geometries of the reference and simulated rasters are different!')
self.categories = referenceMap.getBandGradation(1)
for s in simulatedMap.getBandGradation(1):
if not s in self.categories:
raise EBError('Categories in the reference and simulated rasters are different!')
R = referenceMap.getBand(1)
S = simulatedMap.getBand(1)
self.shape = R.shape
R, S = masks_identity(R,S, dtype=np.uint8)
# Array for weight
self.W = np.ones(self.shape)
self.W = self.W - np.ma.getmask(R)
R = np.ma.filled(R, 0)
S = np.ma.filled(S, 0)
# Proportion of category j in pixel n at the beginning resolution of the reference map
self.Rj = {}
for j in self.categories:
self.Rj[j] = 1.0*binaryzation(R, [j])
# Proportion of category j in pixel n at the beginning resolution of the simulated map
self.Sj = {}
for j in self.categories:
self.Sj[j] = 1.0*binaryzation(S, [j])
def test_binarization(self):
t = binaryzation(self.X, [0,2])
answer1 = np.array([
[False, True, False,],
[False, True, False,],
[True, False,True ]
])
answer2 = np.array([
[False, True, False,],
[False, True, False,],
[True, False,True ]
])
assert_array_equal(t, answer1)
assert_array_equal(t, answer2)
mask = [[False, False, False],
[False, False, False],
[True, False, False]]
data = [[False, True, False],
[False, True, False],
[False, False, True]]
assert_array_equal(binaryzation(np.array(data), [True]), np.ma.array(data=data, mask = mask))
data = np.ma.array(data=data, mask=mask)
assert_array_equal(binaryzation(data, [True]), np.ma.array(data=data, mask = mask))
def _predict(self, state):
'''
Predict the changes.
'''
geodata = state.getGeodata()
rows, cols = geodata['ySize'], geodata['xSize']
# Get locations where self.initStateNum is occurs
band = state.getBand(1)
initStateMask = binaryzation(band, [self.initStateNum])
mask = band.mask
# Calculate summary map of factors weights
# Confidence:
# confidence is summary map of factors, if current state = self.initState
# confidence is 0, if current state != self.initState
# Prediction:
# predicted value is a constant = self.finalStateNum, if current state = self.initState
# predicted value is current state, if current state != self.initState
confidence = np.zeros((rows,cols))
weights = self.getWeights()
weightNum = 0 # Number of processed weights
for f in self.factors:
if not f.geoDataMatch(state):
raise MCEError('Geometries of the state and factor rasters are different!')
f.normalize(mode = 'maxmin')
for i in xrange(f.getBandsCount()):
band = f.getBand(i+1)
confidence = confidence + band*weights[weightNum]
mask = np.ma.mask_or(mask, band.mask)
weightNum = weightNum + 1
confidence = confidence*initStateMask
prediction = np.copy(state.getBand(1))
prediction = np.logical_not(initStateMask) * prediction
prediction = prediction + initStateMask*self.finalStateNum
predicted_band = np.ma.array(data=prediction, mask=mask)
self.prediction = Raster()
self.prediction.create([predicted_band], geodata)
confidence_band = np.ma.array(data=confidence, mask=mask)
self.confidence = Raster()
self.confidence.create([confidence_band], geodata)