k_fold_num = 10
filename = './public/'
# load data: load all the words in all the emails
mailWords, classLables = naiveBayes.loadMailData(filename)
skf = StratifiedKFold(classLables, k_fold_num)
acc_per_fold = []
f1_per_fold = []
recall_per_fold = []
precision_per_fold = []
for train_index, test_index in skf:
print("train_index->", train_index)
print("test_index->", test_index)
preVocabularyList = naiveBayes.createVocabularyList(
[mailWords[i] for i in train_index])
# do wfo filter
vocabularyList = naiveBayes.wfoFilter(
preVocabularyList, [mailWords[i] for i in train_index],
[classLables[i] for i in train_index])
print("vocabularyList finished")
trainMarkedWords = naiveBayes.setOfWordsListToVecTor(
vocabularyList, [mailWords[i] for i in train_index])
print("trainMarkedWords finished")
testMarkedWords = naiveBayes.setOfWordsListToVecTor(
vocabularyList, [mailWords[i] for i in test_index])
# # change it to array
# trainMarkedWords = np.array(trainMarkedWords)
def trainingAdaboostGetDS(iterateNum=40):
test_index = [
2, 6, 7, 8, 13, 16, 19, 29, 35, 37, 40, 42, 43, 45, 46, 49, 51, 52, 64,
65, 71, 72, 78, 79, 80, 84, 85, 90, 91, 98, 103, 109, 111, 117, 123,
129, 135, 138, 142, 149, 169, 188, 191, 192, 203, 221, 225, 226, 229,
232, 236, 243, 250, 254, 257, 258, 259, 264, 268, 281, 298, 300, 308,
319, 322, 329, 333, 335, 338, 339, 340, 344, 347, 358, 359, 362, 382,
385, 391, 394, 402, 410, 415, 417, 418, 422, 423, 424, 425, 428, 437,
441, 456, 461, 462, 470, 472, 477, 480, 481
]
beginTime = datetime.datetime.now()
filename = './public/'
# load data: load all the words in all the emails
mailWords, classLables = naiveBayes.loadMailData(filename)
preVocabularyList = naiveBayes.createVocabularyList(mailWords)
# do wfo filter
vocabularyList = naiveBayes.wfoFilter(preVocabularyList, mailWords,
classLables)
print("length of vocabularyList", len(vocabularyList))
trainMarkedWords = naiveBayes.setOfWordsListToVecTor(
vocabularyList, mailWords)
print("trainMarkedWords finished")
# change it to array
trainMarkedWords = np.array(trainMarkedWords)
print("data to matrix finished")
# calculate each propabilaty of spam and ham P(wi/s) p(wi/h)
pWordsSpamicity, pWordsHealthy, pSpam = \
naiveBayes.trainingNaiveBayes(trainMarkedWords, classLables)
DS = np.ones(len(vocabularyList))
ds_result = {}
minErrorRate = np.inf
for i in range(iterateNum):
errorCount = 0.0
for j in test_index:
testWordsCount = naiveBayes.setOfWordsToVecTor(
vocabularyList, mailWords[j])
testWordsMarkedArray = np.array(testWordsCount)
ps, ph, mailType = naiveBayes.adaboostClassify(
vocabularyList, pWordsSpamicity, pWordsHealthy, DS, pSpam,
testWordsMarkedArray)
if mailType != classLables[j]:
errorCount += 1
alpha = ps - ph
if alpha > 0: # actual: ham; predict:spam
DS[testWordsMarkedArray != 0] = np.abs(
(DS[testWordsMarkedArray != 0] - np.exp(alpha)) /
DS[testWordsMarkedArray != 0])
else: # actual: spam; predict: ham
DS[testWordsMarkedArray != 0] = (
DS[testWordsMarkedArray != 0] +
np.exp(alpha)) / DS[testWordsMarkedArray != 0]
print('DS:', DS)
errorRate = errorCount / len(mailWords)
if errorRate < minErrorRate:
minErrorRate = errorRate
ds_result['minErrorRate'] = minErrorRate
ds_result['DS'] = DS
print('# %d，errorcount %d ，errorrate %f' % (i, errorCount, errorRate))
if errorRate == 0.0:
break
ds_result['vocabularyList'] = vocabularyList
ds_result['pWordsSpamicity'] = pWordsSpamicity
ds_result['pWordsHealthy'] = pWordsHealthy
ds_result['pSpam'] = pSpam
return ds_result
import numpy as np
import datetime
import simpleNavie as naiveBayes
beginTime = datetime.datetime.now()
filename = './public/'
#load data: load all the words in all the emails
smsWords, classLables = naiveBayes.loadMailData(filename)
#get the non-repeated features(vacabulary)
preVocabularyList = naiveBayes.createVocabularyList(smsWords)
#do wfo filter
vocabularyList = naiveBayes.wfoFilter(preVocabularyList, smsWords, classLables)
print("length of vocabularyList", len(vocabularyList))
fw = open('vocabularyList.txt', 'w')
for i in vocabularyList:
fw.write(i + '\n')
fw.flush()
fw.close()
print( "vocabularyList finished")
#change to vector: each email is a vector included in the trainMakredWords
trainMarkedWords = naiveBayes.setOfWordsListToVecTor(vocabularyList, smsWords)
print( "trainMarkedWords finished")
# change it to array
trainMarkedWords = np.array(trainMarkedWords)
print( "data to matrix finished")
#calculate each propabilaty of spam and ham P(wi/s) p(wi/h)
pWordsSpamicity, pWordsHealthy, pSpam = naiveBayes.trainingNaiveBayes(trainMarkedWords, classLables)
print("length of pWordsSpamicity:", len(pWordsSpamicity))
from sklearn import datasets
from sklearn.naive_bayes import GaussianNB
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier
from sklearn.svm import LinearSVC
from sklearn.calibration import calibration_curve, CalibratedClassifierCV
from sklearn.metrics import (brier_score_loss, precision_score, recall_score,
f1_score, log_loss)
from sklearn.cross_validation import train_test_split
import simpleNavie as naiveBayes
filename = './public/'
# load data: load all the words in all the emails
mailWords, classLables = naiveBayes.loadMailData(filename)
preVocabularyList = naiveBayes.createVocabularyList(mailWords)
# do wfo filter
vocabularyList = naiveBayes.wfoFilter(preVocabularyList, mailWords, classLables)
print("vocabularyList finished")
trainMarkedWords = naiveBayes.setOfWordsListToVecTor(vocabularyList, mailWords)
print("trainMarkedWords finished")
test_index = [2, 6, 7]
X_train = [trainMarkedWords[i] for i in range(len(trainMarkedWords))]
X_test = [trainMarkedWords[i] for i in test_index]
y_train = [classLables[i] for i in range(len(trainMarkedWords))]
y_test = [classLables[i] for i in test_index]
lr = LogisticRegression()
def crossValidateEvaluate():
beginTime = datetime.datetime.now()
filename = './public/'
# load data: load all the words in all the emails
mailWords, classLables = naiveBayes.loadMailData(filename)
skf = StratifiedKFold(classLables, k_fold_num)
acc_per_fold = []
f1_per_fold = []
recall_per_fold = []
precision_per_fold = []
for train_index, test_index in skf:
print("train_index->", train_index)
print("test_index->", test_index)
preVocabularyList = naiveBayes.createVocabularyList(
[mailWords[i] for i in train_index])
#do wfo filter
vocabularyList = naiveBayes.wfoFilter(
preVocabularyList, [mailWords[i] for i in train_index],
[classLables[i] for i in train_index])
vocabularyList = preVocabularyList
print("length of vocabularyList", len(vocabularyList))
fw = open('vocabularyList.txt', 'w')
for i in vocabularyList:
fw.write(i + '\n')
fw.flush()
fw.close()
print("vocabularyList finished")
trainMarkedWords = naiveBayes.setOfWordsListToVecTor(
vocabularyList, [mailWords[i] for i in train_index])
print("trainMarkedWords finished")
# change it to array
trainMarkedWords = np.array(trainMarkedWords)
print("data to matrix finished")
# calculate each propabilaty of spam and ham P(wi/s) p(wi/h)
pWordsSpamicity, pWordsHealthy, pSpam = \
naiveBayes.trainingNaiveBayes(trainMarkedWords, [classLables[i] for i in train_index])
fpSpam = open('pSpam.txt', 'w')
spam = pSpam.__str__()
fpSpam.write(spam)
fpSpam.close()
np.savetxt('pWordsSpamicity.txt', pWordsSpamicity, delimiter='\t')
np.savetxt('pWordsHealthy.txt', pWordsHealthy, delimiter='\t')
predict = naiveBayes.predict([mailWords[i] for i in test_index])
#predict = naiveBayes.adaboostPredict([smsWords[i] for i in test_index])
acc_per_fold.append(
accuracy_score([classLables[i] for i in test_index], predict))
f1_per_fold.append(
f1_score([classLables[i] for i in test_index], predict))
recall_per_fold.append(
recall_score([classLables[i] for i in test_index], predict))
precision_per_fold.append(
precision_score([classLables[i] for i in test_index], predict))
print("acc_per_fold:", acc_per_fold)
print("f1_per_fold:", f1_per_fold)
print("recall_per_fold:", recall_per_fold)
print("precision_per_fold:", precision_per_fold)
print("acc_per_fold:", acc_per_fold)
print("f1_per_fold:", f1_per_fold)
print("recall_per_fold:", recall_per_fold)
print("precision_per_fold:", precision_per_fold)
print("k-fold:", k_fold_num, " spend:",
(datetime.datetime.now() - beginTime))