def validateModel(self, printToScreen=False):
""" take the validation data and test it on the network at any given point.
returns a single numeric error value which describes the goodness of the model"""
if NoneType in (type(self.validationFeatures),
type(self.validationLabels), type(self.meta)):
raise Exception("Specify validation features and labels")
predictedOutput = list()
for loopCounter, (rowIndex, row) in enumerate(
self.validationFeatures.iterrows()):
predictedOutput.append([item for item in self.predict(row)])
predictedOutput = pd.DataFrame(
np.array(predictedOutput),
columns=[self.meta.categoricalLabelColumns])
#convert predictions back into a human readable format
predictedMatrix = DataCleaner.deNormalize(
DataCleaner.categoricalToNominal(
self.validationFeatures.join(predictedOutput), self.meta),
self.meta)
expectedMatrix = DataCleaner.deNormalize(
DataCleaner.categoricalToNominal(
self.validationFeatures.join(self.validationLabels),
self.meta), self.meta)
error = comparePrediction(predictedMatrix, expectedMatrix, self.meta,
printToScreen)
return error
def validateModel(self, printToScreen=False):
""" take the validation data and test it on the network at any given point.
returns a single numeric error value which describes the goodness of the model"""
if NoneType in (type(self.validationFeatures), type(self.validationLabels), type(self.meta)):
raise Exception("Specify validation features and labels")
predictedOutput = list()
for loopCounter, (rowIndex, row) in enumerate(self.validationFeatures.iterrows()):
predictedOutput.append([item for item in self.predict(row)])
predictedOutput = pd.DataFrame(np.array(predictedOutput), columns=[self.meta.categoricalLabelColumns])
#convert predictions back into a human readable format
predictedMatrix = DataCleaner.deNormalize(DataCleaner.categoricalToNominal(self.validationFeatures.join(predictedOutput), self.meta), self.meta)
expectedMatrix = DataCleaner.deNormalize(DataCleaner.categoricalToNominal(self.validationFeatures.join(self.validationLabels), self.meta), self.meta)
error = comparePrediction(predictedMatrix, expectedMatrix, self.meta, printToScreen)
return error
li("Dataset has Rows: {0}, Columns: {1}".format(len(data), len(meta.names())))
if args["command"] in ("info"):
sys.exit(0)
if not args["labels"]:
labels = raw_input("Which columns to use as labels? [{0}-{1}]: ".format(1, len(meta.names())))
else:
labels = args["labels"]
setLabels(data, meta, labels)
if len(meta.labelColumns) < 1:
raise Exception("Specify atleast 1 label column")
li("Label Columns: {0}".format(meta.labelColumns))
li("Cleaning Data")
data = DataCleaner.impute(data, meta)
data = DataCleaner.normalize(data, meta)
data = DataCleaner.nominalToCategorical(data, meta)
if args["command"] in ("cross_validate", "train"):
if args["command"] == "cross_validate":
li("Starting {0}-fold cross validation".format(args["folds"]))
error = crossValidate(
data,
meta,
folds=args["folds"],
topology=args["topo"],
iterations=args["epochs"],
weights=args["weights"],
graph=args["graph"],
)
data, meta = loader.load()
li(meta)
li("Dataset has Rows: {0}, Columns: {1}".format(len(data), len(meta.names())))
if args['command'] in ("info"):
sys.exit(0)
if not args['labels']:
labels = raw_input('Which columns to use as labels? [{0}-{1}]: '.format(1, len(meta.names())))
else:
labels = args['labels']
setLabels(data, meta, labels)
if len(meta.labelColumns) < 1:
raise Exception("Specify atleast 1 label column")
li("Label Columns: {0}".format(meta.labelColumns))
li("Cleaning Data")
data = DataCleaner.impute(data, meta)
data = DataCleaner.normalize(data, meta)
data = DataCleaner.nominalToCategorical(data, meta)
if args['command'] in ("cross_validate", "train"):
if args['command'] == "cross_validate":
li("Starting {0}-fold cross validation".format(args['folds']))
error = crossValidate(data, meta, folds=args['folds'], topology=args['topo'], iterations=args['epochs'], weights=args['weights'], graph=args['graph'])
#uncomment the lines if you want to log all files processed
#from os.path import expanduser
#with open("{0}/nncli-history.txt".format(expanduser("~")), "a") as myfile:
#myfile.write("\n{0},{1},{2},{3}".format(args['data'], args['command'], args['topo'], error))
