def PandasDataFrametoDF(df):
'''Convert Pandas DataFrame to DF struct.'''
if df is None:
raise RuntimeError("PandasDataFrametoDF() empty DataFrame")
# Here is a fundamental problem/incompatability between cppEDM and pyEDM
# cppEDM DataFrame stores time vector as strings, and data as valarray
# time values are not strictly required DataFrame( noTime = true )
# Here, we don't have a way to know if the Pandas dataframe passed in
# will have a time vector or not... So... We will just require/assume
# that the first column is ALWAYS a time or index vector.
timeName = df.columns[0]
timeVec = df.get(timeName)
time = [str(x) for x in timeVec] # convert to list of strings
dataList = []
# Add time series data, Skipping the first column!!!
for column in df.columns[1:]:
dataList.append((column, df.get(column).tolist()))
# cppEDM DF struct
DF = pyBindEDM.DF()
DF.timeName = timeName
DF.time = time
DF.dataList = dataList
return DF
def Embed(pathIn="./",
dataFile="",
dataFrame=None,
E=0,
tau=1,
columns="",
verbose=False):
'''Takens time-delay embedding on path/file.
Embed DataFrame columns (subset) in E dimensions.
Calls MakeBlock() after validation and column subset selection.'''
# Establish DF as empty list or Pandas DataFrame for Embed()
if dataFile:
DF = pyBindEDM.DF()
elif isinstance(dataFrame, DataFrame):
if dataFrame.empty:
raise Exception("Embed(): dataFrame is empty.")
DF = pyEDM.AuxFunc.PandasDataFrametoDF(dataFrame)
else:
raise Exception("Embed(): Invalid data input.")
# D is a Python dict from pybind11 < cppEDM Embed
D = pyBindEDM.Embed(pathIn, dataFile, DF, E, tau, columns, verbose)
df = DataFrame(D) # Convert to pandas DataFrame
return df
def Multiview(pathIn="./",
dataFile="",
dataFrame=None,
pathOut="./",
predictFile="",
lib="",
pred="",
D=0,
E=1,
Tp=1,
knn=0,
tau=-1,
columns="",
target="",
multiview=0,
exclusionRadius=0,
trainLib=True,
excludeTarget=False,
verbose=False,
numThreads=4,
showPlot=False):
'''Multiview prediction on path/file.'''
# Establish DF as empty list or Pandas DataFrame for Multiview()
if dataFile:
DF = pyBindEDM.DF()
elif isinstance(dataFrame, DataFrame):
if dataFrame.empty:
raise Exception("Multiview(): dataFrame is empty.")
DF = pyEDM.AuxFunc.PandasDataFrametoDF(dataFrame)
else:
raise Exception("Multiview(): Invalid data input.")
# If lib, pred, columns are not string, but iterable, convert to string
if pyEDM.AuxFunc.NotStringIterable(lib):
lib = ' '.join(map(str, lib))
if pyEDM.AuxFunc.NotStringIterable(pred):
pred = ' '.join(map(str, pred))
if pyEDM.AuxFunc.NotStringIterable(columns):
columns = ' '.join(map(str, columns))
# D is a Python dict from pybind11 < cppEDM Multiview:
# { "View" : < vector< string >, "Predictions" : {} }
D = pyBindEDM.Multiview(pathIn, dataFile, DF, pathOut, predictFile, lib,
pred, D, E, Tp, knn, tau, columns, target,
multiview, exclusionRadius, trainLib,
excludeTarget, verbose, numThreads)
df_pred = DataFrame(D['Predictions']) # Convert to pandas DataFrame
view = DataFrame(D['View'])
if showPlot:
pyEDM.AuxFunc.PlotObsPred(df_pred, dataFile, E, Tp)
MV = {'Predictions': df_pred, 'View': view}
return MV
def PredictNonlinear(pathIn="./",
dataFile="",
dataFrame=None,
pathOut="./",
predictFile="",
lib="",
pred="",
theta="",
E=1,
Tp=1,
knn=0,
tau=-1,
columns="",
target="",
embedded=False,
verbose=False,
numThreads=4,
showPlot=True):
'''Estimate S-map localisation over theta.'''
# Establish DF as empty list or Pandas DataFrame for PredictNonlinear()
if dataFile:
DF = pyBindEDM.DF()
elif isinstance(dataFrame, DataFrame):
if dataFrame.empty:
raise Exception("PredictNonlinear(): dataFrame is empty.")
DF = pyEDM.AuxFunc.PandasDataFrametoDF(dataFrame)
else:
raise Exception("PredictNonlinear(): Invalid data input.")
# If lib, pred, columns are not string, but iterable, convert to string
if pyEDM.AuxFunc.NotStringIterable(lib):
lib = ' '.join(map(str, lib))
if pyEDM.AuxFunc.NotStringIterable(pred):
pred = ' '.join(map(str, pred))
if pyEDM.AuxFunc.NotStringIterable(columns):
columns = ' '.join(map(str, columns))
# D is a Python dict from pybind11 < cppEDM PredictNonlinear
D = pyBindEDM.PredictNonlinear(pathIn, dataFile, DF, pathOut, predictFile,
lib, pred, theta, E, Tp, knn, tau, columns,
target, embedded, verbose, numThreads)
df = DataFrame(D) # Convert to pandas DataFrame
if showPlot:
title = dataFile + "\nE=" + str(E)
ax = df.plot('Theta', 'rho', title=title, linewidth=3)
ax.set(xlabel="S-map Localisation (θ)", ylabel="Prediction Skill ρ")
show()
return df
def EmbedDimension(pathIn="./",
dataFile="",
dataFrame=None,
pathOut="./",
predictFile="",
lib="",
pred="",
maxE=10,
Tp=1,
tau=-1,
columns="",
target="",
embedded=False,
verbose=False,
numThreads=4,
showPlot=True):
'''Estimate optimal embedding dimension [1:maxE].'''
# Establish DF as empty list or Pandas DataFrame for EmbedDimension()
if dataFile:
DF = pyBindEDM.DF()
elif isinstance(dataFrame, DataFrame):
if dataFrame.empty:
raise Exception("EmbedDimension(): dataFrame is empty.")
DF = pyEDM.AuxFunc.PandasDataFrametoDF(dataFrame)
else:
raise Exception("EmbedDimension(): Invalid data input.")
# If lib, pred, columns are not string, but iterable, convert to string
if pyEDM.AuxFunc.NotStringIterable(lib):
lib = ' '.join(map(str, lib))
if pyEDM.AuxFunc.NotStringIterable(pred):
pred = ' '.join(map(str, pred))
if pyEDM.AuxFunc.NotStringIterable(columns):
columns = ' '.join(map(str, columns))
# D is a Python dict from pybind11 < cppEDM CCM
D = pyBindEDM.EmbedDimension(pathIn, dataFile, DF, pathOut, predictFile,
lib, pred, maxE, Tp, tau, columns, target,
embedded, verbose, numThreads)
df = DataFrame(D) # Convert to pandas DataFrame
if showPlot:
title = dataFile + "\nTp=" + str(Tp)
ax = df.plot('E', 'rho', title=title, linewidth=3)
ax.set(xlabel="Embedding Dimension", ylabel="Prediction Skill ρ")
show()
return df
def SMap(pathIn="./",
dataFile="",
dataFrame=None,
pathOut="./",
predictFile="",
lib="",
pred="",
E=0,
Tp=1,
knn=0,
tau=1,
theta=0,
exclusionRadius=0,
columns="",
target="",
smapFile="",
jacobians="",
embedded=False,
verbose=False,
const_pred=False,
showPlot=False):
'''S-Map prediction on path/file.'''
# Establish DF as empty list or Pandas DataFrame for SMap()
if dataFile:
DF = pyBindEDM.DF()
elif isinstance(dataFrame, DataFrame):
if dataFrame.empty:
raise Exception("SMap(): dataFrame is empty.")
DF = pyEDM.AuxFunc.PandasDataFrametoDF(dataFrame)
else:
raise Exception("SMap(): Invalid data input.")
# D is a Python dict from pybind11 < cppEDM SMap:
# { "predictions" : {}, "coefficients" : {} }
D = pyBindEDM.SMap(pathIn, dataFile, DF, pathOut, predictFile, lib, pred,
E, Tp, knn, tau, theta, exclusionRadius, columns,
target, smapFile, jacobians, embedded, const_pred,
verbose)
df_pred = DataFrame(D['predictions']) # Convert to pandas DataFrame
df_coef = DataFrame(D['coefficients']) # Convert to pandas DataFrame
if showPlot:
pyEDM.AuxFunc.PlotObsPred(df_pred, dataFile, E, Tp, False)
pyEDM.AuxFunc.PlotCoeff(df_coef, dataFile, E, Tp)
SMapDict = {'predictions': df_pred, 'coefficients': df_coef}
return SMapDict
def Simplex(pathIn="./",
dataFile="",
dataFrame=None,
pathOut="./",
predictFile="",
lib="",
pred="",
E=0,
Tp=1,
knn=0,
tau=-1,
exclusionRadius=0,
columns="",
target="",
embedded=False,
verbose=False,
const_pred=False,
showPlot=False):
'''Simplex prediction on path/file.'''
# Establish DF as empty list or Pandas DataFrame for Simplex()
if dataFile:
DF = pyBindEDM.DF()
elif isinstance(dataFrame, DataFrame):
if dataFrame.empty:
raise Exception("Simplex(): dataFrame is empty.")
DF = pyEDM.AuxFunc.PandasDataFrametoDF(dataFrame)
else:
raise Exception("Simplex(): Invalid data input.")
# If lib, pred, columns are not string, but iterable, convert to string
if pyEDM.AuxFunc.NotStringIterable(lib):
lib = ' '.join(map(str, lib))
if pyEDM.AuxFunc.NotStringIterable(pred):
pred = ' '.join(map(str, pred))
if pyEDM.AuxFunc.NotStringIterable(columns):
columns = ' '.join(map(str, columns))
# D is a Python dict from pybind11 < cppEDM Simplex
D = pyBindEDM.Simplex(pathIn, dataFile, DF, pathOut, predictFile, lib,
pred, E, Tp, knn, tau, exclusionRadius, columns,
target, embedded, const_pred, verbose)
df = DataFrame(D) # Convert to pandas DataFrame
if showPlot:
pyEDM.AuxFunc.PlotObsPred(df, dataFile, E, Tp)
return df
def CCM(pathIn="./",
dataFile="",
dataFrame=None,
pathOut="./",
predictFile="",
E=0,
Tp=0,
knn=0,
tau=1,
columns="",
target="",
libSizes="",
sample=0,
random=True,
replacement=False,
seed=0,
verbose=False,
showPlot=False):
'''Convergent Cross Mapping on path/file.'''
# Establish DF as empty list or Pandas DataFrame for CCM()
if dataFile:
DF = pyBindEDM.DF()
elif isinstance(dataFrame, DataFrame):
if dataFrame.empty:
raise Exception("CCM(): dataFrame is empty.")
DF = pyEDM.AuxFunc.PandasDataFrametoDF(dataFrame)
else:
raise Exception("CCM(): Invalid data input.")
# D is a Python dict from pybind11 < cppEDM CCM
D = pyBindEDM.CCM(pathIn, dataFile, DF, pathOut, predictFile, E, Tp, knn,
tau, columns, target, libSizes, sample, random,
replacement, seed, verbose)
df = DataFrame(D) # Convert to pandas DataFrame
if showPlot:
title = dataFile + "\nE=" + str(E)
ax = df.plot('LibSize', [df.columns[1], df.columns[2]],
title=title,
linewidth=3)
ax.set(xlabel="Library Size", ylabel="Correlation ρ")
axhline(y=0, linewidth=1)
show()
return df
def PredictInterval(pathIn="./",
dataFile="",
dataFrame=None,
pathOut="./",
predictFile="",
lib="",
pred="",
maxTp=10,
E=1,
tau=1,
columns="",
target="",
embedded=False,
verbose=False,
numThreads=4,
showPlot=True):
'''Estimate optimal prediction interval [1:maxTp]'''
# Establish DF as empty list or Pandas DataFrame for PredictInterval()
if dataFile:
DF = pyBindEDM.DF()
elif isinstance(dataFrame, DataFrame):
if dataFrame.empty:
raise Exception("PredictInterval(): dataFrame is empty.")
DF = pyEDM.AuxFunc.PandasDataFrametoDF(dataFrame)
else:
raise Exception("PredictInterval(): Invalid data input.")
# D is a Python dict from pybind11 < cppEDM PredictInterval
D = pyBindEDM.PredictInterval(pathIn, dataFile, DF, pathOut, predictFile,
lib, pred, maxTp, E, tau, columns, target,
embedded, verbose, numThreads)
df = DataFrame(D) # Convert to pandas DataFrame
if showPlot:
title = dataFile + "\nE=" + str(E)
ax = df.plot('Tp', 'rho', title=title, linewidth=3)
ax.set(xlabel="Forecast Interval", ylabel="Prediction Skill ρ")
show()
return df
def PandasDataFrametoDF( df ):
'''Convert Pandas DataFrame to DF struct.'''
if df is None :
raise RuntimeError( "PandasDataFrametoDF() empty DataFrame" )
# Here is a fundamental problem/incompatability between cppEDM and pyEDM
# cppEDM DataFrame stores time vector as strings, and data as valarray
# time values are not strictly required DataFrame( noTime = true )
# Here, we don't have a way to know if the Pandas dataframe passed in
# will have a time vector or not... So... We will just require/assume
# that the first column is ALWAYS a time or index vector.
# Validate that at least 2 columns are provided
if df.shape[1] < 2:
raise RuntimeError( "PandasDataFrametoDF() DataFrame must have"
" at least 2 columns. First column is time." )
timeName = df.columns[0]
timeVec = df.get( timeName )
time = [ str( x ) for x in timeVec ] # convert to list of strings
# Also require data homogeneity : all numeric, no mixed-data
# but allow a Time first column that is an object...
# in .dtypes: non-numerics are converted to dtype "object"
if any( df.dtypes[1:] == "object" ) :
print( df.dtypes )
raise RuntimeError( "PandasDataFrametoDF() non-numeric data is not"
" allowed in a DataFrame." )
dataList = []
# Add time series data, Skipping the first column!!!
for column in df.columns[1:] :
dataList.append( ( column, df.get( column ).tolist() ) )
# cppEDM DF struct
DF = pyBindEDM.DF()
DF.timeName = timeName
DF.time = time
DF.dataList = dataList
return DF
def PredictInterval( pathIn = "./",
dataFile = "",
dataFrame = None,
pathOut = "./",
predictFile = "",
lib = "",
pred = "",
maxTp = 10,
E = 1,
tau = -1,
exclusionRadius = 0,
columns = "",
target = "",
embedded = False,
verbose = False,
validLib = [],
numThreads = 4,
showPlot = True ):
'''Estimate optimal prediction interval [1:maxTp]'''
# Establish DF as empty list or Pandas DataFrame for PredictInterval()
if dataFile :
DF = pyBindEDM.DF()
elif isinstance( dataFrame, DataFrame ) :
if dataFrame.empty :
raise Exception( "PredictInterval(): dataFrame is empty." )
DF = pyEDM.AuxFunc.PandasDataFrametoDF( dataFrame )
else :
raise Exception( "PredictInterval(): Invalid data input." )
# If lib, pred, columns are not string, but iterable, convert to string
if pyEDM.AuxFunc.NotStringIterable( lib ) :
lib = ' '.join( map( str, lib ) )
if pyEDM.AuxFunc.NotStringIterable( pred ) :
pred = ' '.join( map( str, pred ) )
if pyEDM.AuxFunc.NotStringIterable( columns ) :
columns = ' '.join( map( str, columns ) )
# D is a Python dict from pybind11 < cppEDM PredictInterval
D = pyBindEDM.PredictInterval( pathIn,
dataFile,
DF,
pathOut,
predictFile,
lib,
pred,
maxTp,
E,
tau,
exclusionRadius,
columns,
target,
embedded,
verbose,
validLib,
numThreads )
df = DataFrame( D ) # Convert to pandas DataFrame
if showPlot :
if embedded :
E = len( columns.split() )
title = dataFile + "\nE=" + str( E )
ax = df.plot( 'Tp', 'rho', title = title, linewidth = 3 )
ax.set( xlabel = "Forecast Interval",
ylabel = "Prediction Skill ρ" )
show()
return df
def CCM( pathIn = "./",
dataFile = "",
dataFrame = None,
pathOut = "./",
predictFile = "",
E = 0,
Tp = 0,
knn = 0,
tau = -1,
exclusionRadius = 0,
columns = "",
target = "",
libSizes = "",
sample = 0,
random = True,
replacement = False,
seed = 0,
embedded = False,
includeData = False,
parameterList = False,
verbose = False,
showPlot = False ) :
'''Convergent Cross Mapping on path/file.'''
# Establish DF as empty list or Pandas DataFrame for CCM()
if dataFile :
DF = pyBindEDM.DF()
elif isinstance( dataFrame, DataFrame ) :
if dataFrame.empty :
raise Exception( "CCM(): dataFrame is empty." )
DF = pyEDM.AuxFunc.PandasDataFrametoDF( dataFrame )
else :
raise Exception( "CCM(): Invalid data input." )
# If columns, libSizes, target are not string, but iterable, convert to string
if pyEDM.AuxFunc.NotStringIterable( columns ) :
columns = ' '.join( map( str, columns ) )
if pyEDM.AuxFunc.NotStringIterable( libSizes ) :
libSizes = ' '.join( map( str, libSizes ) )
if pyEDM.AuxFunc.NotStringIterable( target ) :
target = ' '.join( map( str, target ) )
# D is a Python dict from pybind11 < cppEDM CCM
D = pyBindEDM.CCM( pathIn,
dataFile,
DF,
pathOut,
predictFile,
E,
Tp,
knn,
tau,
exclusionRadius,
columns,
target,
libSizes,
sample,
random,
replacement,
seed,
embedded,
includeData,
parameterList,
verbose )
# D has { "LibMeans" : DF }
# and if includeData has : { PredictStats1 : DF, PredictStats2 : DF }
libMeans = DataFrame( D[ "LibMeans" ] ) # Convert to pandas DataFrame
# If includeData, create dict with means and individual prediction stats
if includeData :
CM = { 'LibMeans' : libMeans,
'PredictStats1' : DataFrame( D[ "PredictStats1" ] ),
'PredictStats2' : DataFrame( D[ "PredictStats2" ] ) }
if parameterList and includeData :
CM[ 'parameters' ] = D[ 'parameters' ]
if showPlot :
title = dataFile + "\nE=" + str(E)
ax = libMeans.plot( 'LibSize',
[ libMeans.columns[1], libMeans.columns[2] ],
title = title, linewidth = 3 )
ax.set( xlabel = "Library Size", ylabel = "Correlation ρ" )
axhline( y = 0, linewidth = 1 )
show()
if includeData :
return CM
else :
return libMeans
def SMap( pathIn = "./",
dataFile = "",
dataFrame = None,
pathOut = "./",
predictFile = "",
lib = "",
pred = "",
E = 0,
Tp = 1,
knn = 0,
tau = -1,
theta = 0,
exclusionRadius = 0,
columns = "",
target = "",
smapFile = "",
jacobians = "",
solver = None,
embedded = False,
verbose = False,
const_pred = False,
showPlot = False,
validLib = [],
generateSteps = 0,
parameterList = False
):
'''S-Map prediction on path/file.'''
# Establish DF as empty list or Pandas DataFrame for SMap()
if dataFile :
DF = pyBindEDM.DF()
elif isinstance( dataFrame, DataFrame ) :
if dataFrame.empty :
raise Exception( "SMap(): dataFrame is empty." )
DF = pyEDM.AuxFunc.PandasDataFrametoDF( dataFrame )
else :
raise Exception( "SMap(): Invalid data input." )
# If lib, pred, columns are not string, but iterable, convert to string
if pyEDM.AuxFunc.NotStringIterable( lib ) :
lib = ' '.join( map( str, lib ) )
if pyEDM.AuxFunc.NotStringIterable( pred ) :
pred = ' '.join( map( str, pred ) )
if pyEDM.AuxFunc.NotStringIterable( columns ) :
columns = ' '.join( map( str, columns ) )
# Validate the solver if one was passed in
if solver :
supportedSolvers = [ 'LinearRegression',
'Ridge', 'Lasso', 'ElasticNet',
'RidgeCV', 'LassoCV', 'ElasticNetCV' ]
if not solver.__class__.__name__ in supportedSolvers :
raise Exception( "SMap(): Invalid solver." )
# D is a Python dict from pybind11 < cppEDM SMap:
# { "predictions" : {}, "coefficients" : {}, ["parameters" : {}] }
D = pyBindEDM.SMap( pathIn,
dataFile,
DF,
pathOut,
predictFile,
lib,
pred,
E,
Tp,
knn,
tau,
theta,
exclusionRadius,
columns,
target,
smapFile,
jacobians,
solver,
embedded,
const_pred,
verbose,
validLib,
generateSteps,
parameterList )
df_pred = DataFrame( D['predictions'] ) # Convert to pandas DataFrame
df_coef = DataFrame( D['coefficients'] ) # Convert to pandas DataFrame
SMapDict = { 'predictions' : df_pred, 'coefficients' : df_coef }
if parameterList :
SMapDict[ 'parameters' ] = D[ 'parameters' ]
if showPlot :
if embedded :
E = len( columns.split() )
pyEDM.AuxFunc.PlotObsPred( df_pred, dataFile, E, Tp, False )
pyEDM.AuxFunc.PlotCoeff ( df_coef, dataFile, E, Tp )
return SMapDict
