loss='binary_crossentropy',
metrics=['accuracy'])
return model
sizes = ['10', '50', '100', '150', '200', '250']
methods = ['MRMR', 'JMI', 'JMIM']
for method in methods:
for size in sizes:
global curr_size
curr_size = int(size)
print size
print method
import time
d = DataSetLoader()
X_train = d.LoadDataSet("A")
y_train = d.LoadDataSetClasses("A")
print X_train.shape
print y_train.shape
#chaipee will fix it later on
y_train = numpy.transpose(y_train)
print y_train.shape
targets = list(y_train)
y_train = []
for i in targets:
#print i
y_train.append(int(i))
#print len(y_train)
#first run indices
indices = joblib.load('datasetA_pickles/selected_indices_' + method +
'.joblib.pkl')
from sklearn.neural_network import MLPClassifier
from sklearn.pipeline import make_pipeline
from sklearn import metrics
from sklearn import preprocessing
from sklearn.ensemble import VotingClassifier
from sklearn.model_selection import LeaveOneOut
sizes = ['10', '50', '100', '150', '200', '250']
methods = ['MRMR', 'JMI', 'JMIM']
for method in methods:
for size in sizes:
print size
print method
import time
d = DataSetLoader()
X_train = d.LoadDataSet("B_train")
y_train = d.LoadDataSetClasses("B_train")
print X_train.shape
print y_train.shape
#chaipee will fix it later on
y_train = numpy.transpose(y_train)
print y_train.shape
targets = list(y_train)
y_train = []
for i in targets:
#print i
y_train.append(int(i))
#print len(y_train)
indices = joblib.load('datasetB_pickles/datasetB' + size + '-' +
method + '.joblib.pkl')
preprocessings = ['Standard','Imputer','Robust','Quantile']
#datasets = ["A","B"]
classifiers = ["MLP","SVM","AdaBoost","DT","RandomForest","ExtraTree"]
dataset = "B"
f=open('mcc/mccResults'+dataset+'.txt','w');
f.write("dataset, size, method, classifier, validationTechnique, mc, timeTaken, extra info");
for classifierName in classifiers:
for method in methods:
for size in sizes:
for preproc in preprocessings:
for validation in validationTechnique:
#print size
#print method
d = DataSetLoader();
X_train= d.LoadDataSet("B_train");
y_train = d.LoadDataSetClasses("B_train");
X_test= d.LoadDataSet("B_test");
y_test = d.LoadDataSetClasses("B_test");
#chaipee will fix it later on
y_train=numpy.transpose(y_train)
print y_train.shape
targets=list(y_train)
y_train=[]
for i in targets:
#print i
y_train.append(int(i))
y_test=numpy.transpose(y_train)
print y_test.shape
methods=['MRMR','JMI','JMIM']
sizes=['10','50','100','150','200','250']
classifiers = ["MLP","SVM","RandomForest","AdaBoost","DT","ExtraTree"]
validationTechniques = ["10FoldCV","LOOCV"],
preps=["Standard","Robust","Quantile","Imputer"]
n_iter_search = 20
#Iterating over each method
for dataset in datasets:
f=open('mcc/mccResults'+dataset+'.txt','a');
f.write("dataset, size, method, classifier, validationTechnique, mc, timeTaken, cv.max, cv.mean, cv.min, cv.std, preprocessing");
print "Dataset = ",dataset
#initiating datasetloader object
d = DataSetLoader();
#loading relevant Data and coresponding labels of dataset A
X_train_full = d.LoadDataSet(dataset+"_train");
y_train = d.LoadDataSetClasses(dataset+"_train");
X_validate_full = d.LoadDataSet(dataset+"_test");
y_validate = d.LoadDataSetClasses(dataset+"_test");
print ("Dimensions of training data and labels:",X_train_full.shape,y_train.shape)
print ("Dimensions of validation data and labels:",X_validate_full.shape,y_validate.shape)
#chaipee will fix it later on
targets=list(numpy.transpose(y_train))
y_train=[]
for i in targets:
y_train.append(int(i))
targets=list(numpy.transpose(y_validate))
results = "";
FTs = eval(Dataset + "_FTs");
FT_Accuracies = eval(Dataset + "_FT_Accuracies")
start_time = time.time();
padding = 0;
#load the dataset
d = DataSetLoader();
#X_train_full = d.LoadDataSet(Dataset+"_train");
#y_train = d.LoadDataSetClasses(Dataset+"_train");
#targets=list(numpy.transpose(y_train))
#y_train=[]
#for i in targets:
# y_train.append(int(i))
X_validate_full = d.LoadDataSet(Dataset + "_test");
y_validate = d.LoadDataSetClasses(Dataset + "_test");
print ("Dimensions of validation data and labels:", X_validate_full.shape, y_validate.shape)
targets = list(numpy.transpose(y_validate))
y_validate = []
if Dataset == "C":
y_validate = numpy.array(targets)
#y_validate[y_validate == 0] = -1
else:
for i in targets:
y_validate.append(int(i))
y_test = y_validate
actuals = ','.join([str(elem) for elem in y_test])
actuals = actuals.replace("\n","").replace("[","").replace("]","").replace(" ",",") #to handle dataset C targets
from MachineSpecificSettings import Settings
import scipy.io
from DataSetLoaderLib import DataSetLoader
from sklearn.neural_network import MLPClassifier
from sklearn import metrics
print("")
print("")
print("")
print("")
targets=numpy.array(joblib.load('DatasetA_ValidationClasses.joblib.pkl'))
d = DataSetLoader();
G = d.LoadDataSet("A");
indices= joblib.load('selected_indicesv2.joblib.pkl')
result=numpy.array(G)[:,indices]
clf = MLPClassifier(activation='logistic',solver='sgd')
import time
start_time=time.time()
scores = cross_val_score(clf, result, targets, cv=10)
end_time=time.time()-start_time
print end_time
for i in scores:
print i
print("Accuracy: %0.2f (+/- %0.2f)" % (scores.mean(), scores.std() * 2))
filename='MLP_k-fold.joblib.pkl'
joblib.dump(clf,filename, compress=9)
methods = ['MRMR'] #,'JMI','JMIM'
sizes = ['10'] #,'50','100','150','200','250'
validationTechniques = ["10FoldCV"] #"LOOCV",
preps = ["Standard", "Robust", "Quantile", "Imputer"]
#Iterating over each method
print("Dataset, prepType, validationTechnique, method, size")
for dataset in datasets:
# f=open('mcc/mccEnsembleResults.txt','w');
# f.write("dataset, size, method, classifier, validationTechnique, mc, timeTaken, cv.max, cv.mean, cv.min, cv.std, preprocessing");
print "Dataset = ", dataset
#initiating datasetloader object
d = DataSetLoader()
#loading relevant Data and coresponding labels of dataset A
X_train = d.LoadDataSet(dataset + "_train")
y_train = d.LoadDataSetClasses(dataset + "_train")
X_test = d.LoadDataSet(dataset + "_test")
y_test = d.LoadDataSetClasses(dataset + "_test")
print("Dimensions of validation data and labels:", X_test.shape,
y_test.shape)
#chaipee will fix it later on
targets = list(numpy.transpose(y_train))
y_train = []
for i in targets:
y_train.append(int(i))
targets = list(numpy.transpose(y_test))
y_test = []
"""
TEST PROGRAM
"""
from GlobalUtils import *
import scipy
from MachineSpecificSettings import Settings
import scipy.io
import numpy
from DataSetLoaderLib import DataSetLoader
import csv
#variables =joblib.load('DatasetA_Validation.joblib.pkl')
#targets =joblib.load('DatasetA_ValidationClasses.joblib.pkl')
d = DataSetLoader();
variables = d.LoadDataSet("A");
targets = d.LoadDataSetClasses("A");
def test_range(start,end,G,targets,half):
try:
print "first"
print start
print end-half
SelectSubSetmRMR(G[:,start:end-half],targets)
except:
return 1
try:
print "second"
print start+half
print end
SelectSubSetmRMR(G[:,start+half:end],targets)
i+=1
values.append(i+self.add_by)
joblib.dump(values,'selected_features/datasetC/selected_indices'+'_'+useMethod+'.joblib.pkl', compress=9)
except:
type, value, traceback = sys.exc_info()
print('Error Occured %s: %s: %s' % (type, value, traceback))
threadLock.release()
print len(values)
print "Exiting " + self.name
return
threads=[]
d = DataSetLoader();
G = d.LoadDataSet("C_train");
targets = d.LoadDataSetClasses("C_train");
#chaipee will fix it later on
#y_train=numpy.transpose(targets)
y_train=numpy.asarray(targets)
print y_train.shape
targets=list(y_train)
y_train=[]
for i in targets:
print i
y_train.append(int(i))
targets = y_train
#targets =numpy.asarray(targets )
#targets = column_or_1d(targets, warn=True)
print "Dataset loaded"
def performTreeletClustering(DatasetName):
saveFreq = 1000
#temp value for i
x = -1
if (not (os.path.isfile("objs.pickle"))):
print "New Start"
d = DataSetLoader()
G = d.LoadDataSet(DatasetName)
F = G
M = []
cacheTopXPerPart = d.CacheTopXPerPart(DatasetName)
corrCalculator = PairwisePearsonCorrelationCalculator()
print "calling corr calculator"
corrMatrix = corrCalculator.CalculateSimilarity(
G, d.GetPartSize(DatasetName), cacheTopXPerPart)
else:
print "continuing from where we left off"
d = DataSetLoader()
G, F, M, x, corrMatrix, cacheTopXPerPart = read()
corrCalculator = PairwisePearsonCorrelationCalculator()
p = F[0, :].size
#because we have already done the previous iteration and loaded that one
#save(G,F,M,i,corrMatrix,cacheTopXPerPart)
i = x + 1
print corrMatrix[0]
while i < p:
#for i in range (x+1, p):
recalc = False
if checkCorr(corrMatrix, p) == 0:
print "ERROR IN CORRMATRIX INDEX"
return 0
#calculating value of p
p = F[0, :].size
print "Value of i is : " + str(i) + " out of " + str(p)
theVectors = corrMatrix[0]
#this is always the max corr so the element we want to process
try:
if (corrMatrix[0][3] == ''):
recalc = False
else:
print corrMatrix[0]
recalc = True
except:
pass
Fa = F[:, theVectors[0]]
Fb = F[:, theVectors[1]]
print "calling generate metagene"
m = generateNewMetaGene(Fa, Fb)
print "calling scipy delete on F"
F = scipy.delete(F, theVectors[1], 1)
if not len(M): #if this is the first meta gene in this matrix
M = m
else:
M = numpy.column_stack(
(m, M)) #include in the meta genes set as well
corrMatrix.pop(0)
corrMatrix = corrCalculator.UpdateSimilarity(corrMatrix, F, list(m),
theVectors[0],
theVectors[1])
F[:, theVectors[0]] = m
if len(
corrMatrix
) <= 0 or recalc == True: #everything after this is potentially incorrect so lets recalculate the matrix
corrMatrix = corrCalculator.CalculateSimilarity(
F, d.GetPartSize(DatasetName), cacheTopXPerPart)
if i % saveFreq == 0:
save(G, F, M, i, corrMatrix, cacheTopXPerPart)
i += 1
F = numpy.column_stack((
G,
M)) #scipy.append(G, M, 1) #define a new expanded featureset F = G U M
return F
