def intersect(self, nums1, nums2):
c3 = ct(nums1)&ct(nums2)
print(c3, ct(nums1),ct(nums2))
ret = []
for ch in c3:
ret += ([ch]*c3[ch])
return ret
def shortestCompletingWord(self, licensePlate, words):
ans = None
license = ct(i.lower() for i in licensePlate if i.isalpha())
for i in words:
tmp = ct(i)
if all(k in tmp and tmp[k] >= v for k, v in license.items()):
ans = ans if ans and len(i) >= len(ans) else i
return ans
def isAnagram(self, s, t):
"""
:type s: str
:type t: str
:rtype: bool
"""
from collections import Counter as ct
sct = ct(s)
tct = ct(t)
return sct == tct
def create_3K_lic_rpt(input_file, output_file, smart_account):
'''Creates a CSV formatted Report of 3K licensing content from a file input
of a CCW-R file export'''
with open(input_file, 'r') as f:
rl = f.readlines()
### Find CCW-R header row to place into a list
header = [i for i in rl if ldos(i)]
###Parse CCW-R lines with any 3x50 SKUs into a list of rows
dev_3x50 = [i for i in rl if is_3x50(i)]
###Parse CCW-R lines for traditional top-level SKU rows
non_C1_dev = [i for i in dev_3x50 if non_C1_3x50(i)]
###Parse CCW-R lines for individual on-box SW upgrade licensing rows
upg_lics = [i for i in dev_3x50 if lic_C1_3x50(i)]
###Parse C1 SKUs for 3Ks less than 12 ports b/c SW licenses appear in top-level
non_24_48_port = [i for i in dev_3x50 if non_24_48_port_C1(i)]
###Concatenate all parsed lists
parsed_ccwr_rows_list = header + non_C1_dev + upg_lics + non_24_48_port
###Perform count of elements in concatenated list and place in dict
devdict = dict(ct([(i.split(','))[0] for i in parsed_ccwr_rows_list][1:]))
###Extract top-level SKUs and convert to list of actual licensing SKU that appear in CSSM.
C3x50 = [((i.split(','))[0][3:11] + '-' + (i.split(','))[0][-1])
for i in parsed_ccwr_rows_list
if i.split(',')[0].startswith('WS-C3')]
C3x50 = C3x50 + [((i.split(','))[0][:12].replace(
(i.split(','))[0][:5], 'C') + '-' + (i.split(','))[0][-1])
for i in parsed_ccwr_rows_list
if i.split(',')[0].startswith('C1-WS')]
C3x50_E = [i.replace(i[-2:], '-S-E') for i in C3x50 if i.endswith('E')]
C3x50_S = [i.replace(i[-2:], '-L-S') for i in C3x50 if i.endswith('S')]
###Extract top-level upgrade license SKUs and convert to list
upg_lics_indiv = [i.split(',')[0] for i in upg_lics]
###Concatenate license lists
total_upg_lics = C3x50_E + C3x50_S + upg_lics_indiv
###Perform count of elements in concatenated list and place in dict
licdict = dict(ct(total_upg_lics))
###Create output file
with open(output_file, 'w') as f:
f.write('Top-Level Device OR License,-----,Count\n')
for i in devdict:
f.write(i + ',-----,' + str(devdict[i]) + '\n')
f.write(4 * '\n')
f.write("LICENSES to be deposited in %s\n\n" % smart_account +
'License,-----,Count\n')
for i in licdict:
f.write(i + ',-----,' + str(licdict[i]) + '\n')
f.write(4 * '\n')
f.write("Full License/Device Breakout from CCW-R\n\n")
for i in parsed_ccwr_rows_list:
f.write(i)
def meu_knn(dados_train, rotulo_train, dados_teste, k, normalizar):
rotulos = []
if normalizar:
# aplica normalização aos dados
dados_teste = normalizacao(dados_teste)
dados_train = normalizacao(dados_train)
for i in range(len(dados_teste)):
distancia_teste_train = []
for j in range(len(dados_train)):
# calcula a distância entre o elemento i de teste e o elemento j de treinamento
distancia_teste_train.append(dist(dados_teste[i], dados_train[j], len(dados_teste[i])))
# ordena a lista de distância e rótulos juntos para que fiquem com os mesmos índices nos
# elementos correspondentes; então, seleciona a lista com os rótulos já ordenados
rotulos_ordenados = [y for x, y in sorted(zip(distancia_teste_train, rotulo_train))]
# salva somente a coluna com os rótulos, que é o valor interessante para o problema
rotulos_ordenados = [row[0] for row in rotulos_ordenados]
# seleciona os k primeiros elementos dos rótulos, calcula a quantidade de ocorrências
# para cada valor e pega o valor mais comum encontrado
rotulos.append(ct(rotulos_ordenados[:k]).most_common(1)[0][0])
return rotulos
def removeDuplicates(data: list) -> list:
output = data[::-1]
count = ct(data)
for i in [k for k, v in count.items() if v > 2]:
for _ in range(count[i] - 2): # 需删除的次数
output.remove(i)
print(output[::-1])
return output[::-1]
def topKFrequent1(self, nums, k):
"""
:type nums: List[int]
:type k: int
:rtype: List[int]
"""
from collections import Counter as ct
return [k for (k,v) in ct(nums).most_common(k)]
def findLucky(self, arr: List[int]) -> int:
C = ct(arr)
ret = -1
for c in C :
if c == C[c] :
ret = max(ret, c)
return ret
def top_publishers(publishers, top):
top_ten = set((""))
top_ten = dict(ct(publishers).most_common(top))
#for publisher, count in publishers.most_common():
print(top_ten)
print(len(top_ten))
return top_ten
def top_Platforms(platforms, top):
top_platforms = set((""))
top_platforms = dict(ct(platforms).most_common(top))
#for publisher, count in publishers.most_common():
print(top_platforms)
print(len(top_platforms))
return top_platforms
def top_publishers(publishers,
top): #deletes publishers with less than min number
top_ten = set((""))
top_ten = dict(ct(publishers).most_common(top))
#for publisher, count in publishers.most_common():
print(top_ten)
print(len(top_ten))
return top_ten
def customSortString(self, S, T):
"""
:type S: str
:type T: str
:rtype: str
"""
t = set(T)
t2 = set(S)
from collections import Counter as ct
c = ct(T)
s = [char * c[char] for char in S if char in t]
add = [char * c[char] for char in t - t2]
return "".join(s + add)
def f(l):
#print l
if len(l) == 1:
return l[0]
m = min(x[0] for x in l)
two = filter(lambda x: x[0] == m, l)
#print two,m
if len(two) == 2:
rest = filter(lambda x: x[0] != m, l)
c = ct(two[0][1:] + two[1][1:])
c -= ct([x[0] for x in rest])
assert len(c) == 1
return list(c.elements()) + f(sorted([x[1:] for x in rest]))
else:
m = max(x[-1] for x in l)
two = filter(lambda x: x[-1] == m, l)
rest = filter(lambda x: x[-1] != m, l)
assert len(two) == 2
c = ct(two[-1][:-1] + two[-2][:-1])
c -= ct([x[-1] for x in rest])
assert len(c) == 1
return f(sorted([x[:-1] for x in rest])) + list(c.elements())
def select_example(self):
sub_pred = dd(list) #Mn predicted labels for each cluster
idx = 0
for k,v in self.ex_id.items():
sub_pred[k] = self.clf.predict(self.fn[v]) #predict labels for cluster learning set
#entropy-based cluster selection
rank = []
for k,v in sub_pred.items():
count = list(ct(v).values())
count[:] = [i/float(max(count)) for i in count]
H = np.sum(-p*math.log(p,2) for p in count if p!=0)
rank.append([k,len(v),H])
rank = sorted(rank, key=lambda x: x[-1], reverse=True)
if not rank:
raise ValueError('no clusters found in this iteration!')
c_idx = rank[0][0] #pick the 1st cluster on the rank, ordered by label entropy
c_ex_id = self.ex_id[c_idx] #examples in the cluster picked
sub_label = sub_pred[c_idx] #used when choosing cluster by H
sub_fn = self.fn[c_ex_id]
#sub-cluster the cluster
c_ = KMeans(init='k-means++', n_clusters=len(np.unique(sub_label)), n_init=10)
c_.fit(sub_fn)
dist = np.sort(c_.transform(sub_fn))
ex_ = dd(list)
for i,j,k,l in zip(c_.labels_, c_ex_id, dist, sub_label):
ex_[i].append([j,l,k[0]])
for i,j in ex_.items(): #sort by ex. dist to the centroid for each C
ex_[i] = sorted(j, key=lambda x: x[-1])
for k,v in ex_.items():
if v[0][0] not in self.labeled_set: #find the first unlabeled ex
idx = v[0][0]
break
return idx, c_idx
def topKFrequent(nums, k):
from collections import Counter as ct
return [num for (num, count) in ct(nums).most_common(k)]
z = 0
for i in teslaSummaryPOS:
teslaSummaryTags.append(teslaSummaryPOS[x][1])
x += 1
for i in NYTimesPOS:
NYTimesTags.append(NYTimesPOS[y][1])
y += 1
for i in ESPNPOS:
ESPNTags.append(ESPNPOS[z][1])
z +=1
# In[92]:
#count the tags
teslaSumTagCount = ct(teslaSummaryTags)
NYTimesTagCount = ct(NYTimesTags)
ESPNTagCount = ct(ESPNTags)
# In[93]:
#sublinear normalization
for key in teslaSumTagCount:
x = teslaSumTagCount[key]
if x > 0:
teslaSubLinearNorm[key] = 1 + math.log(x)
for key in NYTimesTagCount:
x = NYTimesTagCount[key]
if x > 0:
NYTimesSubLinearNorm[key] = 1 + math.log(x)
def count_publisher(json_data):
publishers = ct(k['publisher'] for k in json_data if k.get('publisher'))
# for publisher, count in publishers.most_common():
# print(publisher, count)
return publishers
else:
print '# of p label', len(p_label)
print 'p label acc', sum(label[p_idx]==p_label)/float(len(p_label))
#print 'ex before 30 itr', ct(ex_30)
#print 'ex after 50 itr', ct(ex_50)
#print 'ex all', ct(ex_all)
cm_cls = np.unique(np.hstack((test_label,preds)))
f = open('al_out','w')
f.writelines('%s;\n'%repr(i) for i in tp_type)
f.write('ex in each itr:'+repr(ex)+'\n')
f.write(repr(cm_cls))
f.close()
print 'f count on all ex', ct(label)
ave_acc = [np.mean(acc) for acc in acc_sum]
acc_std = [np.std(acc) for acc in acc_sum]
'''
ave_acc_type = [[] for i in range(6)]
ave_pre = [[] for i in range(6)]
ave_rec = [[] for i in range(6)]
for i in range(6):
ave_acc_type[i] = [np.mean(a) for a in acc_type[i]]
ave_pre[i] = [np.mean(p) for p in precision_type[i]]
ave_rec[i] = [np.mean(r) for r in recall_type[i] ]
'''
print 'overall acc:', repr(ave_acc)
print 'p1',p1
print np.mean(p1)
from collections import Counter as ct
l=["ZERO", "ONE", "TWO", "THREE", "FOUR", "FIVE", "SIX", "SEVEN", "EIGHT", "NINE"]
g=map(str,range(0,10))
def f(a,b):
tp=ct()
for i in b.keys():
tp[i]=a*b[i]
return tp
#l=map(lambda x:ct(x),l)
for i in range(int(raw_input())):
c=ct(raw_input())
d=ct()
ans=[0]*10
ans[0]=c['Z']
ans[2]=c['W']
ans[8]=c['G']
ans[6]=c['X']
ans[4]=c['U']
ans[3]=c['H']-ans[8]
ans[5]=c['F']-ans[4]
ans[7]=c['V']-ans[5]
ans[1]=c['O']-ans[2]-ans[4]-ans[0]
ans[9]=(c['N']-ans[1]-ans[7])/2
#print ans
assert(len(filter(lambda a:a>=0,ans))==10)
assert(ct("".join(map(lambda a,b: a*b,ans,l)))==c)
print "Case #{}: {}".format(str(i+1),"".join(map(lambda a,b:a*b,ans,g)))
label1 = input2[:, -1]
label = input4[:, -1]
label1 = input6[:, -1]
# input3 = input3 #quick run of the code using other building
# input3, label = shuffle(input3, label)
name = []
for i in input3:
s = re.findall("(?i)[a-z]{2,}", i)
name.append(" ".join(s))
cv = CV(analyzer="char_wb", ngram_range=(3, 4))
# tv = TV(analyzer='char_wb', ngram_range=(3,4))
fn = cv.fit_transform(name).toarray()
# fn = cv.fit_transform(input1).toarray()
# print cv.vocabulary_
print "class count of true labels of all ex:\n", ct(label)
# n_class = len(np.unique(label))
# print n_class
# print np.unique(label)
# print 'class count from groud truth labels:\n',ct(label)
fold = 10
rounds = 100
# clf = LR()
# clf = LinearSVC()
# clf = SVC(kernel='linear', probability=True)
clf = RFC(n_estimators=100, criterion="entropy")
# kf = StratifiedKFold(label, n_folds=fold, shuffle=True)
kf = KFold(len(label), n_folds=fold, shuffle=True)
mapping = {
def create_3K_lic_rpt(ccwr_full_list, output_file, smart_account):
'''Create a group on lambda functions to perform RegEx searches'''
###Find Header Row based on Serial keyword.
hr = lambda s: search('.*[Ss][Ee][Rr][Ii][Aa].*', s)
###Find any SKU containing 3K nomenclature.
is_3x50 = lambda s: search('3[68]50', s)
###Find any top level traditonal hardware SKU that also contains license level.
non_C1_3x50 = lambda s: search('WS-C3[68]50.*-[SE]', s)
###Find individual 3K on-box license SKUs.
lic_C1_3x50 = lambda s: search('C3[68]50-[24][48]-[SL]-[ES]', s)
###Find any C1 SKU that is less than 24 ports. These have license level as part of the top-level part.
non_24_48_port_C1 = lambda s: search('C1-WS.*-12.*', s)
'''Creates a CSV formatted Report of 3K licensing content from a file input
of a CCW-R file export'''
#print(dumps(ccwr_full_list[0:5],indent=4))
### Find CCW-R header row to place into a list
header = [i for i in ccwr_full_list[0:3] if hr(str(i))]
###Parse CCW-R lines with any 3x50 SKUs into a list of rows
dev_3x50 = [i for i in ccwr_full_list if is_3x50(str(i))]
###Parse CCW-R lines for traditional top-level SKU rows
non_C1_dev = [i for i in dev_3x50 if non_C1_3x50(str(i))]
#print(dumps(non_C1_dev,indent=4))
###Parse CCW-R lines for individual on-box SW upgrade licensing rows
upg_lics = [i for i in dev_3x50 if lic_C1_3x50(str(i))]
#print(dumps(upg_lics,indent=4))
###Parse C1 SKUs for 3Ks less than 12 ports b/c SW licenses appear in top-level
non_24_48_port = [i for i in dev_3x50 if non_24_48_port_C1(str(i))]
#print(dumps(non_24_48_port,indent=4))
###Concatenate all parsed lists
parsed_ccwr_rows_list = header + non_C1_dev + upg_lics + non_24_48_port
#print(dumps(parsed_ccwr_rows_list,indent=4))
###Perform count of elements in concatenated list and place in dict
devdict = dict(ct([i[0] for i in parsed_ccwr_rows_list][1:]))
###Extract top-level SKUs and convert to list of actual licensing SKU that appear in CSSM.
C3x50 = [
i[0][3:11] + '-' + i[0][-1] for i in parsed_ccwr_rows_list
if i[0].startswith('WS-C3')
]
C3x50 = C3x50 + [
i[0][:12].replace(i[0][:5], 'C' + '-' + i[0][-1])
for i in parsed_ccwr_rows_list if i[0].startswith('C1-WS')
]
C3x50_E = [i.replace(i[-2:], '-S-E') for i in C3x50 if i.endswith('E')]
C3x50_S = [i.replace(i[-2:], '-L-S') for i in C3x50 if i.endswith('S')]
###Extract top-level upgrade license SKUs and convert to list
upg_lics_indiv = [i[0] for i in upg_lics]
###Concatenate license lists
total_upg_lics = C3x50_E + C3x50_S + upg_lics_indiv
###Perform count of elements in concatenated list and place in dict
licdict = dict(ct(total_upg_lics))
###Create output file
with open(output_file, 'w') as f:
f.write('Top-Level Device OR License,-----,Count\n')
for i in devdict:
f.write(i + ',-----,' + str(devdict[i]) + '\n')
f.write(4 * '\n')
f.write("LICENSES to be deposited in %s\n\n" % smart_account +
'License,-----,Count\n')
for i in licdict:
f.write(i + ',-----,' + str(licdict[i]) + '\n')
f.write(4 * '\n')
f.write("Full License/Device Breakout from CCW-R\n\n")
for i in parsed_ccwr_rows_list:
for j in i:
f.write(j)
f.write(',')
f.write('\n')
featureLabelFile = "../../dataset/processed_acl/processedBooksKitchenElectronics/"+dataName
featureMatrix, labelList = readFeatureLabel(featureLabelFile)
# transferLabelFile = "../../dataset/processed_acl/processedBooksKitchenElectronics/transferLabel_books--electronics.txt"
transferLabelFile = "../../dataset/processed_acl/processedBooksKitchenElectronics/transferLabel_kitchen--electronics.txt"
auditorLabelList, transferLabelList, trueLabelList = readTransferLabel(transferLabelFile)
featureMatrix = np.array(featureMatrix)
labelArray = np.array(labelList)
transferLabelArray = np.array(transferLabelList)
print("number of types", len(set(labelArray)))
print('class count of true labels of all ex:\n', ct(transferLabelArray))
initialExList = [[397, 1942, 200], [100, 1978, 657], [902, 788, 1370], [1688, 1676, 873], [1562, 1299, 617], [986, 1376, 562], [818, 501, 1922], [600, 1828, 1622], [1653, 920, 1606], [39, 1501, 166]]
fold = 10
rounds = 150
multipleClassFlag = False
al = _ProactiveLearning(fold, rounds, featureMatrix, labelArray, transferLabelArray, "sentiment_electronics", multipleClassFlag)
al.setInitialExList(initialExList)
al.run_CV()
"""
def run_CV(self, batchSize):
cvIter = 0
totalInstanceNum = len(self.label)
print("totalInstanceNum\t", totalInstanceNum)
indexList = [i for i in range(totalInstanceNum)]
totalTransferNumList = []
random.shuffle(indexList)
foldNum = 10
foldInstanceNum = int(totalInstanceNum * 1.0 / foldNum)
foldInstanceList = []
for foldIndex in range(foldNum - 1):
foldIndexInstanceList = indexList[foldIndex *
foldInstanceNum:(foldIndex + 1) *
foldInstanceNum]
foldInstanceList.append(foldIndexInstanceList)
foldIndexInstanceList = indexList[foldInstanceNum * (foldNum - 1):]
foldInstanceList.append(foldIndexInstanceList)
cvIter = 0
totalAccList = [[] for i in range(10)]
for foldIndex in range(foldNum):
# self.clf = LinearSVC(random_state=3)
# self.clf = LR(fit_intercept=False)
self.clf = LR(multi_class="multinomial", solver='lbfgs')
train = []
for preFoldIndex in range(foldIndex):
train.extend(foldInstanceList[preFoldIndex])
test = foldInstanceList[foldIndex]
for postFoldIndex in range(foldIndex + 1, foldNum):
train.extend(foldInstanceList[postFoldIndex])
trainNum = int(totalInstanceNum * 0.9)
fn_test = self.fn[test]
label_test = self.label[test]
print("testing", ct(label_test))
fn_train = self.fn[train]
initExList = []
initExList = self.pretrainSelectInit(train)
# initExList = [316, 68, 495]
# random.seed(110)
# initExList = random.sample(train, 3)
fn_init = self.fn[initExList]
label_init = self.label[initExList]
print("initExList\t", initExList, label_init)
queryIter = 3
labeledExList = []
unlabeledExList = []
###labeled index
labeledExList.extend(initExList)
unlabeledExList = list(set(train) - set(labeledExList))
while queryIter < rounds:
fn_train_iter = []
label_train_iter = []
fn_train_iter = self.fn[labeledExList]
label_train_iter = self.label[labeledExList]
self.clf.fit(fn_train_iter, label_train_iter)
idxList = self.select_example(unlabeledExList, batchSize)
# print(queryIter, "idx", idx, self.label[idx])
# self.update_select_confidence_bound(idx)
if len(idxList) > 0:
labeledExList += idxList
for idx in idxList:
unlabeledExList.remove(idx)
else:
labeledExList.append(idxList)
unlabeledExList.remove(idxList)
acc = self.get_pred_acc(fn_test, label_test, labeledExList)
totalAccList[cvIter].append(acc)
queryIter += 1
cvIter += 1
totalACCFile = modelVersion + "_acc.txt"
f = open(totalACCFile, "w")
for i in range(10):
totalAlNum = len(totalAccList[i])
for j in range(totalAlNum):
f.write(str(totalAccList[i][j]) + "\t")
f.write("\n")
f.close()
def parentheses(s):
c = ct(s)
if c['('] == c[')']:
return True
return False
from sklearn import tree
from sklearn.preprocessing import normalize
input1 = [i.strip().split('+')[-1][:-5] for i in open('sdh_pt_rice').readlines()]
input21 = np.genfromtxt('keti_hour_sum', delimiter=',')
input22 = np.genfromtxt('sdh_hour_rice', delimiter=',')
input2 = np.vstack((input21,input22))
#input2 = np.genfromtxt('sdh_45min_forrice', delimiter=',')
input3 = [i.strip().split('\\')[-1][:-5] for i in open('rice_pt_sdh').readlines()]
input4 = np.genfromtxt('rice_hour_sdh', delimiter=',')
input5 = [i.strip().split('_')[-1][:-5] for i in open('soda_pt_new').readlines()]
input6 = np.genfromtxt('soda_45min_new', delimiter=',')
label1 = input2[:,-1]
label = input4[:,-1]
label1 = input6[:,-1]
print 'class count of true labels of all ex:\n', ct(label)
#input3 = input3 #quick run of the code using other building
name = []
for i in input3:
s = re.findall('(?i)[a-z]{2,}',i)
name.append(' '.join(s))
cv = CV(analyzer='char_wb', ngram_range=(3,4))
#tv = TV(analyzer='char_wb', ngram_range=(3,4))
fn = cv.fit_transform(name).toarray()
#fn = cv.fit_transform(input1).toarray()
#print cv.vocabulary_
#fd = input4[:,[0,1,2,3,5,6,7]]
#kmer = cv.get_feature_names()
#idf = zip(kmer, cv._tfidf.idf_)
#idf = sorted(idf, key=lambda x: x[-1], reverse=True)
def f(a,b):
tp=ct()
for i in b.keys():
tp[i]=a*b[i]
return tp
featureMatrix.append(featureList)
if line[lineLen-1] == "FALSE":
label.append(0.0)
else:
# print(line[lineLen-1])
label.append(1.0)
f.close()
return featureMatrix, label
if __name__ == "__main__":
# featureLabelFile = "../../dataset/processed_acl/processedKitchenElectronics/"+dataName
featureLabelFile = "../../dataset/processed_acl/processedBooksElectronics/"+dataName
featureMatrix, labelList = readFeatureLabel(featureLabelFile)
featureMatrix = np.array(featureMatrix)
labelArray = np.array(labelList)
print('class count of true labels of all ex:\n', ct(labelArray))
transferLabelFile = "../../dataset/processed_acl/processedBooksElectronics/transferLabel_books--electronics.txt"
auditorLabelList, transferLabelList, targetLabelList = readTransferLabel(transferLabelFile)
transferLabelArray = np.array(transferLabelList)
fold = 10
rounds = 100
al = active_learning(fold, rounds, featureMatrix, transferLabelArray, labelArray)
al.run_CV()
#print 'ct on traing label', ct(train_label)
clf.fit(train_fn, train_label)
sub_pred = dd(list) #Mn predicted labels for each cluster
for k,v in ex_id.items():
sub_pred[k] = clf.predict(fn[v]) #predict labels for cluster learning set
#acc_ = accuracy_score(label[train_], preds_c)
#print 'acc on test set', acc
#print 'acc on cluster set', acc_
#acc_sum[rr].append(acc)
#print 'iteration', rr, '>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>'
#the original H based cluster selection
rank = []
for k,v in sub_pred.items():
count = ct(v).values()
count[:] = [i/float(max(count)) for i in count]
H = np.sum(-p*math.log(p,2) for p in count if p!=0)
#H /= len(v)/float(len(train))
rank.append([k,len(v),H])
rank = sorted(rank, key=lambda x: x[-1], reverse=True)
if not rank:
break
idx = rank[0][0] #pick the id of the 1st cluster on the rank
cl_id.append(idx) #track cluster id on each iteration
cc = idx #id of the cluster picked by H
c_id = ex_id[cc] #example id of the cluster picked
sub_label = sub_pred[idx]#used when choosing cluster by H
sub_fn = fn[c_id]
#sub-clustering the cluster
gnb.fit(train[newarr].values, train["Patient Condition"])
result = gnb.predict(test[newarr])
# Print Performance Indicator
print(
"Number of mislabeled points out of a total {} points : {}, performance {:05.2f}%"
.format(
test.shape[0], (test["Patient Condition"] != result).sum(), 100 *
(1 - (test["Patient Condition"] != result).sum() / test.shape[0])))
test_data = pd.concat([test[newarr], test["Patient Condition"]], axis=1)
test_data["Patient Condition"] = result
print(test_data)
counts = ct(result)
count_p = counts['Positive']
count_n = counts['Negative']
slices = [count_p, count_n]
cols = ['b', 'c']
plt.pie(slices,
labels=['Positve', 'Negative'],
colors=cols,
shadow=True,
startangle=90,
autopct='%1.1f%%')
plt.title("Patient Condition")
plt.legend()
plt.show()
ora_idx.append(idx)
#print k,label_md[idx],label[idx],cf_md[idx],input1[idx]
#elif cf_md[v[itr][0]] <=0.2:
#print '>>>>>',k,label_md[idx],label[idx],cf_md[idx],input1[idx]
ex_ora.append(len(ora_idx))
'''
train_data = data1[train]
train_label = label1[train]
validate_data = data1[validate]
validate_label = label1[validate]
'''
train_data = fn[np.hstack((auto_idx,ora_idx))]
train_label = np.hstack((label_md[auto_idx],label[ora_idx]))
train_label_ = label[np.hstack((auto_idx,ora_idx))]
print ct(train_label)
clf.fit(train_data,train_label)
acc = clf.score(test_data,test_label)
#acc_sum[itr].append(acc)
acc_H.append(acc)
clf.fit(train_data,train_label_)
acc = clf.score(test_data,test_label)
acc_T.append(acc)
'''
cm = CM(test_label,preds)
cm = normalize(cm.astype(np.float), axis=1, norm='l1')
k=0
while k<clx:
acc_type[k].append(cm[k,k])
k += 1
y = 0
z = 0
for i in teslaSummaryPOS:
teslaSummaryTags.append(teslaSummaryPOS[x][1])
x += 1
for i in NYTimesPOS:
NYTimesTags.append(NYTimesPOS[y][1])
y += 1
for i in ESPNPOS:
ESPNTags.append(ESPNPOS[z][1])
z += 1
# In[92]:
#count the tags
teslaSumTagCount = ct(teslaSummaryTags)
NYTimesTagCount = ct(NYTimesTags)
ESPNTagCount = ct(ESPNTags)
# In[93]:
#sublinear normalization
for key in teslaSumTagCount:
x = teslaSumTagCount[key]
if x > 0:
teslaSubLinearNorm[key] = 1 + math.log(x)
for key in NYTimesTagCount:
x = NYTimesTagCount[key]
if x > 0:
NYTimesSubLinearNorm[key] = 1 + math.log(x)
for key in ESPNTagCount:
def func_p():
x = 1
file_name = sd.askstring("File Name", "Enter your file name ")
data = pd.read_csv(file_name)
while x == 1:
if file_name is not None:
x = 0
si("", "File Loaded!")
#mapping data
data["Sex_cleaned"] = np.where(
data["Sex"] == "M", "Male",
(np.where(data["Sex"] == "F", "Female", "Infant")))
data = data[[
"Sex_cleaned", "Length", "Diameter", "Height", "Whole weight",
"Shucked weight", "Viscera weight", "Shell weight", "Rings"
]].dropna(axis=0, how='any')
#Split dataset
train, test = train_test_split(data,
test_size=0.6,
random_state=int(4))
gnb = GaussianNB()
indicators = [
"Length", "Diameter", "Height", "Whole weight",
"Shucked weight", "Viscera weight", "Shell weight", "Rings"
]
gnb.fit(train[indicators].values, train["Sex_cleaned"])
y_pred = gnb.predict(test[indicators])
#Print Performance Indicator
data_accuracy = (
"Total data {} points : {}, Accuracy {:05.2f}%".format(
test.shape[0], (test["Sex_cleaned"] != y_pred).sum(), 100 *
(1 -
(test["Sex_cleaned"] != y_pred).sum() / test.shape[0])))
test_data = pd.concat([test[indicators], test["Sex_cleaned"]],
axis=1)
test_data["Sex Prediction"] = y_pred
test_data["Data Accuracy"] = data_accuracy
#Excel Writer
writer = pd.ExcelWriter('Output.xlsx', engine='xlsxwriter')
test_data.to_excel(writer, sheet_name='Sheet1')
writer.save()
si("", "Output Created! Check it out!")
answer = mb.askyesno("Question",
"Do you want to check the Data Chart?")
if answer == True:
#Counter
counts = ct(y_pred)
count_Male = counts['Male']
count_Female = counts['Female']
count_Infant = counts['Infant']
slices = [count_Male, count_Female, count_Infant]
cols = ['#00ffc3', '#ff00cb', '#ffd000']
#Pie Chart
fig = Figure(figsize=(100, 100))
a = fig.add_subplot(111)
a.pie(slices,
labels=['Male', 'Female', 'Infant'],
colors=cols,
shadow=True,
startangle=90,
autopct='%1.1f%%')
a.legend()
canvas = FigureCanvasTkAgg(fig)
canvas.get_tk_widget().pack(fill=BOTH, expand=True)
return fig
FigureCanvasTk.draw()
else:
root.destroy()
else:
sw("Warning", "Please re - enter your input")
def parentheses(s):
c = ct(s)
if c["("] == c[")"]:
return True
return False
j += 1
k += 1
if k == 512:
listaFrequenciaY.append(j)
k = 0
j = 0
# calcula frequencias Z
for i in range(len(freqIdaZ)):
if freqIdaZ[i] == freqVoltaZ[i]:
j += 1
k += 1
if k == 512:
listaFrequenciaZ.append(j)
k = 0
j = 0
contagemFrequenciaX = ct(listaFrequenciaX)
media = 0.0
variancia = 0.0
desvio = 0.0
qtd = 0.0
eventoFrequencia = []
#print resultados
print(ninho + '-' + tratamento)
print(
'Freq X - media: ' +
str(format(calcMedia(listaFrequenciaX), '.2f')),
'desvio: ' + str(format(calcDesvio(listaFrequenciaX), '.2f')),
'min: ' + str(min(listaFrequenciaX)),
'max: ' + str(max(listaFrequenciaX)))
print(
'Freq Y - media: ' +
featureLabelFile = "../../dataset/processed_acl/processedBooksElectronics/" + dataName
featureMatrix, labelList = readFeatureLabel(featureLabelFile)
featureMatrix = np.array(featureMatrix)
labelArray = np.array(labelList)
###processedKitchenElectronics transferLabel_electronics--kitchen.txt
transferLabelFile = "../../dataset/processed_acl/processedBooksElectronics/transferLabel_books--electronics.txt"
auditorLabelList, transferLabelList, targetLabelList = readTransferLabel(
transferLabelFile)
transferLabelArray = np.array(transferLabelList)
auditorLabelArray = np.array(auditorLabelList)
# print(auditorLabel)
# exit()
# label = np.array([float(i.strip()) for i in open('targetAuditorLabel.txt').readlines()])
# tmp = np.genfromtxt('../../data/rice_hour_sdh', delimiter=',')
# label = tmp[:,-1]
print('class count of true labels of all ex:\n', ct(labelArray))
print("count of auditor", ct(auditorLabelArray))
# exit()
# mapping = {1:'co2',2:'humidity',4:'rmt',5:'status',6:'stpt',7:'flow',8:'HW sup',9:'HW ret',10:'CW sup',11:'CW ret',12:'SAT',13:'RAT',17:'MAT',18:'C enter',19:'C leave',21:'occu'}
# fn = get_name_features(raw_pt)
fold = 10
rounds = 100
al = active_learning(fold, rounds, featureMatrix, auditorLabelArray)
al.run_CV()
f.close()
return featureMatrix, label
if __name__ == "__main__":
###processedKitchenElectronics electronics ---> kitchen
###processedBooksElectronics books ---> electronics
sourceFeatureLabelFile = "../../dataset/processed_acl/processedBooksElectronics/"+sourceDataName
sourceFeatureMatrix, sourceLabelList = readFeatureLabel(sourceFeatureLabelFile)
sourceLabel = np.array(sourceLabelList)
sourceFeatureMatrix = np.array(sourceFeatureMatrix)
print('class count of true source labels of all ex:\n', ct(sourceLabel))
targetFeatureLabelFile = "../../dataset/processed_acl/processedBooksElectronics/"+targetDataName
targetFeatureMatrix, targetLabelList = readFeatureLabel(targetFeatureLabelFile)
targetLabel = np.array(targetLabelList)
targetFeatureMatrix = np.array(targetFeatureMatrix)
print('class count of true target labels of all ex:\n', ct(targetLabel))
fold = 1
rounds = 100
al = active_learning(fold, rounds, sourceFeatureMatrix, sourceLabel, targetFeatureMatrix, targetLabel)
dataDir = "../../dataset/processed_acl/processedBooksElectronics/"
def count_Platform(json_data):
platforms = ct(k.platform for k in json_data if k.platform)
# for publisher, count in publishers.most_common():
# print(publisher, count)
return platforms
label2 = input2[:,-1]
label = input4[:,-1]
#input3, label = shuffle(input3, label)
name = []
for i in input3:
s = re.findall('(?i)[a-z]{2,}',i)
name.append(' '.join(s))
vc = CV(analyzer='char_wb', ngram_range=(3,4), min_df=1, token_pattern='[a-z]{2,}')
#vc = TV(analyzer='char_wb', ngram_range=(3,4), min_df=1, token_pattern='[a-z]{2,}')
fn = vc.fit_transform(name).toarray()
fd = input4[:,[0,1,2,3,5,6,7]]
#n_class = len(np.unique(label))
#print n_class
#print np.unique(label)
print ct(label)
#kmer = vc.get_feature_names()
#idf = zip(kmer, vc._tfidf.idf_)
#idf = sorted(idf, key=lambda x: x[-1], reverse=True)
#print idf[:20]
#print idf[-20:]
#print vc.get_feature_names()
fold = 2
kf = StratifiedKFold(label, n_folds=fold, shuffle=True)
#kf = KFold(len(label), n_folds=fold, shuffle=True)
'''
folds = [[] for i in range(fold)]
i = 0
for train, test in kf:
folds[i] = test
