def main():
parser = argparse.ArgumentParser(description='Generates train.r. A sample command is mGenForE.py -e ob/e1/ ')
parser.add_argument('-e', required=True,help='Experiement folder to use to find the features and targets')
parser.add_argument('-a', required=True,help='Algorithm name')
parser.add_argument('-s', required=True,help='Location of the folder containing all the sub experiments')
parser.add_argument('-targetClass',required=True,help="binomial(target takes only true and false) / multinomial (target values takes more than 2 values)")
parser.add_argument('-skipM',required=False,help="yes or no , If you want to regenerate already generated algorithm model file then make this value No")
parser.add_argument('-td',required=True,help="Day on which it was trained")
parser.add_argument('-dt',required=True,help="Number of days it was trained")
parser.add_argument('-wt',required=True,help="default/exp , weight type to be given to different days")
parser.add_argument('-iT',required=False,help='Instrument name')
parser.add_argument('-sP',required=False,help='Strike price of instrument')
parser.add_argument('-oT',required=False,help='Options Type')
args = parser.parse_args()
attribute.initializeInstDetails(args.iT,args.sP,args.oT)
if args.skipM == None:
args.skipM = "yes"
print "Using the experiment folder " + args.e
config = ConfigObj(args.e+"/design.ini")
print "The config parameters that I am working with are"
print config
dirName=os.path.dirname(args.e)
algo = rCodeGen.getAlgoName(args)
args.s = args.s + "/"
rProgName = "train" + algo + "-td." + os.path.basename(os.path.abspath(args.td)) + "-dt." + args.dt + "-wt." + args.wt+ attribute.generateExtension() +\
"-For"+os.path.basename(os.path.dirname(args.s))+"SubE.r"
rProgLocation = dirName+'/'+rProgName
rScript = open(rProgLocation,'w')
rScript.write('#!/usr/bin/Rscript \n')
if(algo == 'glmnet'):
rScript.write('require (glmnet) \n')
elif(algo == 'randomForest'):
rScript.write('require (randomForest) \n')
elif(algo == 'mda'):
rScript.write('require (mda) \n')
rCodeGen.ForSetUpChecks(rScript)
rCodeGen.ToReadTargetFile(rScript,config)
rCodeGen.ForWtVectorGeneration(rScript,args.wt.lower())
for target in config['target']:
rCodeGen.ToReadFeatureFiles(rScript,config,target)
rCodeGen.ForSanityChecks(rScript,config,target)
designFiles = utility.list_files(args.s)
for designFile in designFiles:
print "Generating r code for " + designFile
rScript.write('\n\nprint ("Running r code for ' + designFile + '")\n')
config = ConfigObj(designFile)
for target in config['target']:
lModelGeneratedAfterTraining = os.path.dirname(designFile) + '/' + algo + target + '-td.' + os.path.basename(os.path.abspath(args.td)) + \
'-dt.' + args.dt + '-targetClass.' + args.targetClass + "-wt." + args.wt+ attribute.generateExtension() + '.model'
if os.path.isfile(lModelGeneratedAfterTraining)and ( args.skipM.lower() == "yes" ):
print "Model File " + lModelGeneratedAfterTraining + " already exists . So it will not be formed again . If you want to re-generate model then re-run with -skipM=No"
else:
rCodeGen.ToCreateDataFrameForTraining(rScript,config,target)
rCodeGen.ForTraining(rScript,args,config,target)
rCodeGen.saveTrainingModel(rScript,args,os.path.dirname(designFile),target)
rScript.write('rm(list=ls())')
rScript.close()
print "Finished generating R training program: " + rProgLocation
os.system("chmod +x "+rProgLocation)
def main():
parser = argparse.ArgumentParser(description='Generates train.r. A sample command is mGenForE.py -e ob/e1/ ')
parser.add_argument('-e', required=True,help='Experiement folder to use to find the features and targets')
parser.add_argument('-a', required=True,help='Algorithm name')
parser.add_argument('-targetClass',required=True,help="binomial(target takes only true and false) / multinomial (target values takes more than 2 values)")
parser.add_argument('-skipT',required=False,help="yes or no , If you want to regenerate already generated algorithm model file then make this value No")
parser.add_argument('-td',required=True,help="Day on which it was trained")
parser.add_argument('-dt',required=True,help="Number of days it was trained")
parser.add_argument('-wt',required=True,help="default/exp , weight type to be given to different days")
parser.add_argument('-iT',required=False,help='Instrument name')
parser.add_argument('-sP',required=False,help='Strike price of instrument')
parser.add_argument('-oT',required=False,help='Options Type')
parser.add_argument('-treeType',required=False,help="Tree read for trade engine")
parser.add_argument('-tTD',required=False,help="Tree number of days to be used")
args = parser.parse_args()
attribute.initializeInstDetails(args.iT,args.sP,args.oT)
if args.skipT == None:
args.skipT = "yes"
if args.tTD == None:
args.tTD = args.dt
print "Using the experiment folder " + args.e
config = ConfigObj(args.e+"/design1.ini")
configInit = ConfigObj(args.e+"design.ini")
# configInitList = []
# for iniFile in os.listdir(args.e + "/"):
# if '.ini' in iniFile and iniFile != 'design.ini':
# index = iniFile[ file.index(".") - 1 ]
# configInitList.append( ( index, ConfigObj(args.e+"/"+iniFile) ) )
# configInit = dict(configInitList)
print "The config parameters that I am working with are"
print config
dirName=os.path.dirname(args.e)+"/"
algo = rCodeGen.getAlgoName(args)
rProgName = "traintree" + "-td." + os.path.basename(os.path.abspath(args.td)) + "-tTD" + args.tTD + "-dt." + args.dt + "-wt." + args.wt + attribute.generateExtension() +".r"
rProgLocation = dirName+'/'+rProgName
rScript = open(rProgLocation,'w')
rScript.write('#!/usr/bin/Rscript \n')
rScript.write('require (rpart) \n')
rCodeGen.ForSetUpChecks(rScript)
lAllFilePresent = True
for target in config['target']:
lTreeFileName = dirName+"/"+algo+ target+'-td.' + os.path.basename(os.path.abspath(args.td)) + "-tTD" + args.tTD + '-dt.' + args.dt + attribute.generateExtension() +".tree" + args.treeType
if os.path.isfile(lTreeFileName) and ( args.skipT.lower() == "yes" ):
continue
else:
lAllFilePresent = False
break
if lAllFilePresent == False:
rCodeGen.ToReadTargetFile(rScript,config)
rCodeGen.ForWtVectorGeneration(rScript,args.wt.lower())
for target in config['target']:
rCodeGen.ToReadFeatureFiles(rScript,config,target)
rCodeGen.ToReadPredictionFiles(rScript,config,target,configInit)
rCodeGen.ForSanityChecks(rScript,config,target)
lTreeFileName = dirName+"/"+algo+ target+'-td.' + os.path.basename(os.path.abspath(args.td)) + "-tTD" + args.tTD +'-dt.' + args.dt + attribute.generateExtension() +".tree" + args.treeType
if os.path.isfile(lTreeFileName) and ( args.skipT.lower() == "yes" ):
print "Model File " + lTreeFileName + " already exists . So it will not be formed again . If you want to re-generate model then re-run with -skipT=No"
else:
rCodeGen.ToRenameDataBeforeTraining(rScript,config,target)
rCodeGen.ForTrainingTree(rScript,args,config,target, args.treeType)
print lTreeFileName
rCodeGen.saveTrainingTree(rScript,args,dirName,target, lTreeFileName)
rScript.close()
print "Finished generating R training program: " + rProgLocation
os.system("chmod +x "+rProgLocation)
def main():
parser = argparse.ArgumentParser(description='Generates train.r. A sample command is mGenForE.py -e ob/e1/ ')
parser.add_argument('-e', required=True,help='Experiement folder to use to find the features and targets')
parser.add_argument('-a', required=True,help='Algorithm name')
parser.add_argument('-s', required=True,help='Location of the folder containing all the sub experiments')
parser.add_argument('-targetClass',required=True,help="binomial(target takes only true and false) / multinomial (target values takes more than 2 values)")
parser.add_argument('-skipM',required=False,help="yes or no , If you want to regenerate already generated algorithm model file then make this value No")
parser.add_argument('-skipP',required=False,help="yes or no , If you want to regenerate already generated algorithm prediction file then make this value No")
parser.add_argument('-pd', required=True,help='Prediction directory')
parser.add_argument('-td',required=True,help="Day on which it was trained")
parser.add_argument('-dt',required=True,help="Number of days it was trained")
parser.add_argument('-wt',required=True,help="default/exp , weight type to be given to different days")
parser.add_argument('-iT',required=False,help='Instrument name')
parser.add_argument('-sP',required=False,help='Strike price of instrument')
parser.add_argument('-oT',required=False,help='Options Type')
args = parser.parse_args()
attribute.initializeInstDetails(args.iT,args.sP,args.oT)
if args.skipM == None:
args.skipM = "yes"
if args.skipP == None:
args.skipP = "yes"
print "Using the experiment folder " + args.e
print "Training files steps"
config = ConfigObj(args.e+"/design.ini")
print "The config parameters that I am working with are"
print config
predictionDataDirectoryName = args.pd.replace('/ro/','/wf/')
predictionDataDirectoryName = predictionDataDirectoryName + "/p/" + os.path.basename(os.path.dirname(args.e))
if not os.path.exists(predictionDataDirectoryName):
os.mkdir(predictionDataDirectoryName)
dirName=os.path.dirname(args.e)
algo = rCodeGen.getAlgoName(args)
args.s = args.s + "/"
rProgName = "trainPredict"+ algo + "-td." + os.path.basename(os.path.abspath(args.td)) + "-dt." + args.dt +\
"-pd." + os.path.basename(os.path.abspath(args.pd)) + "-wt." + args.wt+ attribute.generateExtension() + "-For"+os.path.basename(os.path.dirname(args.s))+"SubE.r"
rProgLocation = dirName+'/'+rProgName
rScript = open(rProgLocation,'w')
rScript.write('#!/usr/bin/Rscript \n')
if(algo == 'glmnet'):
rScript.write('require (glmnet) \n')
elif(algo == 'randomForest'):
rScript.write('require (randomForest) \n')
elif(algo == 'mda'):
rScript.write('require (mda) \n')
rCodeGen.ForSetUpChecksForTrainPredictTogather(rScript)
rCodeGen.ToReadTargetFile(rScript,config)
rCodeGen.ForWtVectorGeneration(rScript,args.wt.lower())
for target in config['target']:
rCodeGen.ToReadFeatureFiles(rScript,config,target,2)
rCodeGen.ForSanityChecks(rScript,config,target)
print "For prediction data set"
configForPredictions = ConfigObj(args.e+"/design.ini")
print "The config parameters that I am working with are"
for target in configForPredictions['target']:
feature_keys = configForPredictions['features-'+target].keys()
features = configForPredictions['features-'+target]
for key in feature_keys:
new_key = key + "P"
features[new_key] = features[key]
del features[key]
print configForPredictions
rCodeGen.ToReadFeatureFiles(rScript,configForPredictions,target,4)
rCodeGen.ForSanityChecks(rScript,configForPredictions,target)
designFiles = utility.list_files(args.s)
for designFile in designFiles:
print "Generating r code for " + designFile
rScript.write('\n\nprint ("Running r code for ' + designFile + '")\n')
config = ConfigObj(designFile)
configForPredictions = ConfigObj(designFile)
#--------------MODEL--------------------
for target in config['target']:
feature_keys = configForPredictions['features-'+target].keys()
features = configForPredictions['features-'+target]
for key in feature_keys:
new_key = key + "P"
features[new_key] = features[key]
del features[key]
lModelGeneratedAfterTraining = os.path.dirname(designFile) + '/' + algo + target + '-td.' + os.path.basename(os.path.abspath(args.td)) + \
'-dt.' + args.dt + '-targetClass.' + args.targetClass + "-wt." + args.wt+ attribute.generateExtension() + '.model'
if os.path.isfile(lModelGeneratedAfterTraining)and ( args.skipM.lower() == "yes" ):
print "Model File " + lModelGeneratedAfterTraining + " already exists . So it will not be formed again . If you want to re-generate model then re-run with -skipM=No"
else:
rCodeGen.ToCreateDataFrameForTraining(rScript,config,target)
rCodeGen.ForTraining(rScript,args,config,target)
rCodeGen.saveTrainingModel(rScript,args,os.path.dirname(designFile),target)
#--------------Prediction Part--------------------
predictionFileName = predictionDataDirectoryName + "/" + args.a + target + '-td.' + os.path.basename(os.path.abspath(args.td)) + \
'-dt.' + args.dt + '-targetClass.' + args.targetClass + '-f.' + os.path.basename(os.path.dirname(designFile)) +\
"-wt." + args.wt+ attribute.generateExtension() +".predictions"
if not os.path.isfile(predictionFileName) or ( args.skipP.lower() == "no" ):
rCodeGen.ForPredictions(rScript,configForPredictions,args,designFile,target,4)
else:
print "Prediction File " + predictionFileName + "Already exists , not generating it again . If you want to generate it again then rerun it with -skipP no "
rScript.write('rm(list=ls())')
rScript.close()
print "Finished generating R training program: " + rProgLocation
os.system("chmod +x "+rProgLocation)
def main():
parser = argparse.ArgumentParser(
description=
'Generates train.r. A sample command is mGenForE.py -e ob/e1/ ')
parser.add_argument(
'-e',
required=True,
help='Experiement folder to use to find the features and targets')
parser.add_argument('-a', required=True, help='Algorithm name')
parser.add_argument(
'-targetClass',
required=True,
help=
"binomial(target takes only true and false) / multinomial (target values takes more than 2 values)"
)
parser.add_argument(
'-skipT',
required=False,
help=
"yes or no , If you want to regenerate already generated algorithm model file then make this value No"
)
parser.add_argument('-td',
required=True,
help="Day on which it was trained")
parser.add_argument('-dt',
required=True,
help="Number of days it was trained")
parser.add_argument(
'-wt',
required=True,
help="default/exp , weight type to be given to different days")
parser.add_argument('-iT', required=False, help='Instrument name')
parser.add_argument('-sP',
required=False,
help='Strike price of instrument')
parser.add_argument('-oT', required=False, help='Options Type')
parser.add_argument('-treeType',
required=False,
help="Tree read for trade engine")
parser.add_argument('-tTD',
required=False,
help="Tree number of days to be used")
args = parser.parse_args()
attribute.initializeInstDetails(args.iT, args.sP, args.oT)
if args.skipT == None:
args.skipT = "yes"
if args.tTD == None:
args.tTD = args.dt
print "Using the experiment folder " + args.e
config = ConfigObj(args.e + "/design1.ini")
configInit = ConfigObj(args.e + "design.ini")
# configInitList = []
# for iniFile in os.listdir(args.e + "/"):
# if '.ini' in iniFile and iniFile != 'design.ini':
# index = iniFile[ file.index(".") - 1 ]
# configInitList.append( ( index, ConfigObj(args.e+"/"+iniFile) ) )
# configInit = dict(configInitList)
print "The config parameters that I am working with are"
print config
dirName = os.path.dirname(args.e) + "/"
algo = rCodeGen.getAlgoName(args)
rProgName = "traintree" + "-td." + os.path.basename(
os.path.abspath(args.td)
) + "-tTD" + args.tTD + "-dt." + args.dt + "-wt." + args.wt + attribute.generateExtension(
) + ".r"
rProgLocation = dirName + '/' + rProgName
rScript = open(rProgLocation, 'w')
rScript.write('#!/usr/bin/Rscript \n')
rScript.write('require (rpart) \n')
rCodeGen.ForSetUpChecks(rScript)
lAllFilePresent = True
for target in config['target']:
lTreeFileName = dirName + "/" + algo + target + '-td.' + os.path.basename(
os.path.abspath(args.td)
) + "-tTD" + args.tTD + '-dt.' + args.dt + attribute.generateExtension(
) + ".tree" + args.treeType
if os.path.isfile(lTreeFileName) and (args.skipT.lower() == "yes"):
continue
else:
lAllFilePresent = False
break
if lAllFilePresent == False:
rCodeGen.ToReadTargetFile(rScript, config)
rCodeGen.ForWtVectorGeneration(rScript, args.wt.lower())
for target in config['target']:
rCodeGen.ToReadFeatureFiles(rScript, config, target)
rCodeGen.ToReadPredictionFiles(rScript, config, target, configInit)
rCodeGen.ForSanityChecks(rScript, config, target)
lTreeFileName = dirName + "/" + algo + target + '-td.' + os.path.basename(
os.path.abspath(args.td)
) + "-tTD" + args.tTD + '-dt.' + args.dt + attribute.generateExtension(
) + ".tree" + args.treeType
if os.path.isfile(lTreeFileName) and (args.skipT.lower() == "yes"):
print "Model File " + lTreeFileName + " already exists . So it will not be formed again . If you want to re-generate model then re-run with -skipT=No"
else:
rCodeGen.ToRenameDataBeforeTraining(rScript, config, target)
rCodeGen.ForTrainingTree(rScript, args, config, target,
args.treeType)
print lTreeFileName
rCodeGen.saveTrainingTree(rScript, args, dirName, target,
lTreeFileName)
rScript.close()
print "Finished generating R training program: " + rProgLocation
os.system("chmod +x " + rProgLocation)
if args.skipP == None:
args.skipP = "yes"
if args.skipT == None:
args.skipT = "yes"
if args.mpMearge == None:
args.mpMearge = "yes"
if args.dt == None:
args.dt = "1"
if args.treeUsed == None:
args.treeUsed = "no"
if(args.sequence == "dp"):
import dp
if args.skipTr == None:
args.skipTr = "yes"
algo = rCodeGen.getAlgoName(args)
if args.a is not None:
allAlgos = [args.a]
else:
allAlgos = ['logitr','glmnet','randomForest']
config = ConfigObj(args.e+"/design.ini")
targetAttributes = attribute.getTargetVariableKeys(config)
one_feature_attributes = attribute.getFeatureVariableKeys(config , targetAttributes.keys()[0])
lengthOfFeatures = len(one_feature_attributes)
allDataDirectories = attribute.getListOfTrainingDirectoriesNames( int(args.nDays) , args.td ,args.iT)
dataFolder = args.td
generatorsFolder = args.g
commandList = []
def main():
parser = argparse.ArgumentParser(description='Generates train.r. A sample command is mGenForE.py -e ob/e1/ ')
parser.add_argument('-e', required=True,help='Experiement folder to use to find the features and targets')
parser.add_argument('-a', required=True,help='Algorithm name')
parser.add_argument('-targetClass',required=True,help="binomial(target takes only true and false) / multinomial (target values takes more than 2 values)")
parser.add_argument('-skipM',required=False,help="yes or no , If you want to regenerate already generated algorithm model file then make this value No")
parser.add_argument('-td',required=True,help="Day on which it was trained")
parser.add_argument('-dt',required=True,help="Number of days it was trained")
parser.add_argument('-wt',required=True,help="default/exp , weight type to be given to different days")
parser.add_argument('-iT',required=False,help='Instrument name')
parser.add_argument('-sP',required=False,help='Strike price of instrument')
parser.add_argument('-oT',required=False,help='Options Type')
parser.add_argument('-double',required=False,help='Double training of in model')
args = parser.parse_args()
attribute.initializeInstDetails(args.iT,args.sP,args.oT)
if args.skipM == None:
args.skipM = "yes"
print "Using the experiment folder " + args.e
config = ConfigObj(args.e+"/design.ini")
print "The config parameters that I am working with are"
print config
dirName=os.path.dirname(args.e)+"/"
algo = rCodeGen.getAlgoName(args)
if args.double:
rProgName = "train" + algo + "-td." + os.path.basename(os.path.abspath(args.td)) + "-dt." + args.dt + "-wt." + args.wt + attribute.generateExtension() +"double.r"
else:
rProgName = "train" + algo + "-td." + os.path.basename(os.path.abspath(args.td)) + "-dt." + args.dt + "-wt." + args.wt + attribute.generateExtension() +".r"
rProgLocation = dirName+'/'+rProgName
rScript = open(rProgLocation,'w')
rScript.write('#!/usr/bin/Rscript \n')
if(algo == 'glmnet'):
rScript.write('require (glmnet) \n')
elif(algo == 'randomForest'):
rScript.write('require (randomForest) \n')
elif(algo == 'mda'):
rScript.write('require (mda) \n')
rCodeGen.ForSetUpChecks(rScript)
lAllFilePresent = True
for target in config['target']:
if args.double:
lModelGeneratedAfterTraining = dirName + '/' + algo + target + '-td.' + os.path.basename(os.path.abspath(args.td))\
+ '-dt.' + args.dt + '-targetClass.' + args.targetClass + "-wt." + args.wt+ attribute.generateExtension() +'double.model'
else:
lModelGeneratedAfterTraining = dirName + '/' + algo + target + '-td.' + os.path.basename(os.path.abspath(args.td))\
+ '-dt.' + args.dt + '-targetClass.' + args.targetClass + "-wt." + args.wt+ attribute.generateExtension() +'.model'
if os.path.isfile(lModelGeneratedAfterTraining) and ( args.skipM.lower() == "yes" ):
continue
else:
lAllFilePresent = False
break
if lAllFilePresent == False:
rCodeGen.ToReadTargetFile(rScript,config)
rCodeGen.ForWtVectorGeneration(rScript,args.wt.lower())
for target in config['target']:
rCodeGen.ToReadFeatureFiles(rScript,config,target)
rCodeGen.ForSanityChecks(rScript,config,target)
if args.double:
lModelGeneratedAfterTraining = dirName + '/' + algo + target + '-td.' + os.path.basename(os.path.abspath(args.td))\
+ '-dt.' + args.dt + '-targetClass.' + args.targetClass + "-wt." + args.wt+ attribute.generateExtension() +'double.model'
lTempModelName = dirName + '/' + algo + target + '-td.' + os.path.basename(os.path.abspath(args.td))\
+ '-dt.' + args.dt + '-targetClass.' + args.targetClass + "-wt." + args.wt+ attribute.generateExtension() +'.model'
else:
lModelGeneratedAfterTraining = dirName + '/' + algo + target + '-td.' + os.path.basename(os.path.abspath(args.td))\
+ '-dt.' + args.dt + '-targetClass.' + args.targetClass + "-wt." + args.wt+ attribute.generateExtension() +'.model'
if os.path.isfile(lModelGeneratedAfterTraining) and ( args.skipM.lower() == "yes" ):
print "Model File " + lModelGeneratedAfterTraining + " already exists . So it will not be formed again . So it will not be formed again . If you want to re-generate model then re-run with -skipM=No"
else:
rCodeGen.ToCreateDataFrameForTraining(rScript,config,target)
if args.double:
if os.path.isfile(lTempModelName):
rCodeGen.ForLoadingModel(rScript,args,dirName,target,config)
else:
rCodeGen.ForTraining(rScript,args,config,target)
rCodeGen.forPreparingWtVectorForDoubleTraining(rScript,args,target)
rCodeGen.saveTrainingModel(rScript,args,dirName,target,"double")
else:
rCodeGen.ForTraining(rScript,args,config,target)
rCodeGen.saveTrainingModel(rScript,args,dirName,target)
rScript.close()
print "Finished generating R training program: " + rProgLocation
os.system("chmod +x "+rProgLocation)