def RandomForest_ALL(self):
if not os.path.exists(self.RF_dir):
os.makedirs(self.RF_dir)
for i in range(1,101):
path = "{0}/random{1}".format(self.RF_dir,i)
if not os.path.exists(path):
os.makedirs(path)
self.ReadCombine()
p = MyPool(self.thred_num)
p.map(self.RandomSampling,range(1,101))
p.map(self.RandomForest,range(1,101))
def DataShrink(self,vec_time):
def func(i):
output_dir = "SVM/Training/{0}".format(self.id)
path = "{0}/training_{1}".format(output_dir, i)
train_linker, train_domain = self.DataDivide(path)
random.shuffle(train_domain)
vec_num = vec_time * len(train_linker)
if vec_num > len(train_domain):
print "Vector number is too learge to decrease!"
exit(1)
train_sampled = train_linker + train_domain[:vec_num]
filename = "{0}/training_sampled_{1}_{2}".format(output_dir,vec_time,i)
with open(filename, "w") as fp:
fp.writelines(" ".join(x)+"\n" for x in train_sampled)
p = MyPool(self.thred_num)
p.map(func,range(1,6))
def MakeVectorDataForCross(self):
TASKS = []
for x in ["Training", "Test"]:
dir_name = "{0}/SVM/{1}/{2}".format(self.work_dir,x,self.id)
if not os.path.isdir(dir_name):
os.makedirs(dir_name)
for y in range(1,6):
if x == "Training":
linker_f = "{0}/Crosslist_{1}".format(self.linker_dir, y)
output_f = "{0}/training_{1}".format(dir_name, y)
elif x == "Test":
linker_f = "{0}/crosslist_{1}".format(self.linker_dir, y)
output_f = "{0}/test_{1}".format(dir_name, y)
#make list of arguments
TASKS.append((linker_f, output_f))
linker_f = "{0}/svmcrosslist".format(self.linker_dir)
output_f = "{0}/SVM/Test/{1}/test_all".format(self.work_dir,self.id)
TASKS.append((linker_f,output_f))
#########multiprocessing
p = MyPool(self.thred_num)
p.map(self.MakeVectorData,TASKS)
def SVM_Multi(self):
p = MyPool(self.thred_num)
p.map(self.SVMLearn, range(1,6))
p.map(self.SVMClassify, range(1,6))
command = "cat {0}/output_? > {0}/output_sw1".format(self.output_dir)
os.system(command)
