def tarea1(entrenamiento, prueba):
d = Main()
(t_0, t_1) = d.split(entrenamiento)
nb = NaiveBayes.NaiveBayes(entrenamiento, t_1, t_0, prueba)
nb.plot()
b = Bayes.Bayes(entrenamiento, t_1, t_0, prueba)
b.plot()
return
def test():
labels = []
label_ids = set()
doc_matrix = []
dir = 'data/'
train_file = 'train.txt'
test_file = 'test.txt'
word_set_file = 'all_words.txt'
model_file = 'model.txt'
with open(dir + train_file) as f:
for l in f:
l = l.replace('\n','')
if l == '':
continue
comps = l.split('\t')
assert(len(comps) == 2)
if comps[1] == '':
continue
labels.append(comps[0])
doc_matrix.append(comps[1].split(','))
label_ids.add(comps[0])
bayes_model = Bayes.Bayes(dir + word_set_file)
bayes_model.train(doc_matrix, labels, list(label_ids), dir + model_file)
#open the test file
expect_labels = []
predict_docs = []
with open(dir + test_file) as f:
for l in f:
l = l.replace('\n', '')
if l == '':
continue
comps = l.split('\t')
if comps[1] == '':
continue
assert(len(comps) == 2)
expect_labels.append(comps[0])
predict_docs.append(comps[1].split(','))
predict_labels = bayes_model.predict(dir + model_file, predict_docs)
post_analysis(predict_labels, expect_labels)
from Bayes import *
import commands
import re
print '3b'
bc = Bayes()
bc.train('../data/arxiv/arxiv.train')
bc.predict('../data/arxiv/arxiv.test', 0, 1, 1, 0)
print '3c'
c = Bayes()
c.train('../data/arxiv/arxiv.train')
c.predict('../data/arxiv/arxiv.test', 0, 1, 10, 0)
print '3d'
nfold = 4
s_test = []
s_train = []
for d in range(nfold):
s_test = []
s_train = []
with open('../data/arxiv/arxiv.norm.train', 'r') as f:
for i, l in enumerate(f):
if i % nfold == d:
s_test.append(l)
else:
s_train.append(l)
with open('../data/arxiv/arxiv.norm%d.test' % d, 'w') as test:
for t in s_test:
test.write(t)
print("Dictionary classes created")
print("Creating and completing positive and negative dictionaries...")
if SIZED_DCT is False:
dictionary1.create_dictionary()
print(f"Positive dictionary created")
dictionary0.create_dictionary()
print(f"Negative dictionary created \n \n")
else:
dictionary1.create_sized_dictionary(SIZE)
print("Positive dictionary created")
dictionary0.create_sized_dictionary(SIZE)
print("Negative dictionary created \n \n")
print("Creating BAYES class...")
bayes = Bayes(dictionary1, dictionary0, testing_set)
print("Bayes class created")
print("Predicting sentiments for testing set...")
nb_undetermined = bayes.predict_sentiments(LAPLACE_SMOOTHING,
pos_spl_nb, neg_spl_nb)
print("Prediction of sentiments for testing set done")
print(
f"Number of tweets with undetermined sentiments : {nb_undetermined}"
)
print(
"Comparing sentiments from the dataset with predicted sentiments..."
)
metrics, conf_matrix = bayes.compare_sentiments()
import SVM import lr import Bayes import LDA LDA.LDA() Bayes.Bayes() SVM.svmwch() lr.lr()
def performBayes(inputDataClass,
drawPrecisionRecall=False,
drawConfusion=False):
"""################################# Bayes Classifier #############################################"""
##Sklearn
# print("\nSklearn Naive Bayes")
# clf = GaussianNB()
# clf.fit(inputDataClass.Train[:,:-1], inputDataClass.Train[:,-1])
# Ypred = clf.predict(inputDataClass.Train[:,:-1])
# Ytrue = inputDataClass.Train[:,-1]
# print("Training Accuracy = "+str(performanceAnalyser.calcAccuracyTotal(Ypred,Ytrue)))
# Ypred = clf.predict(inputDataClass.Test[:,:-1])
# Ytrue = inputDataClass.Test[:,-1]
# print("Testing Accuracy = "+str(performanceAnalyser.calcAccuracyTotal(Ypred,Ytrue)))
print("\nMy Naive Bayes")
bayesClassifier = Bayes.Bayes(
isNaive=False,
distribution=[0 for i in range(inputDataClass.Train.shape[1] - 1)])
# bayesClassifier = Bayes.Bayes(isNaive = True, distribution =[-1,0,0,1,1,0])
bayesClassifier.train(inputDataClass.Train)
print("Training of model done.")
Ypred = bayesClassifier.fit(inputDataClass.Train)
Ytrue = inputDataClass.Train[:, -1]
print("Training Accuracy = " +
str(performanceAnalyser.calcAccuracyTotal(Ypred, Ytrue)))
Ypred = bayesClassifier.fit(inputDataClass.Test)
Ytrue = inputDataClass.Test[:, -1]
print("Testing Accuracy = " +
str(performanceAnalyser.calcAccuracyTotal(Ypred, Ytrue)))
print("Prediction done.")
if drawConfusion:
confusion = performanceAnalyser.getConfusionMatrix(Ytrue, Ypred)
Visualization.visualizeConfusion(confusion)
if drawPrecisionRecall:
############################ precision-recall curve #############################
threshold = np.arange(0.9, 0.1, -0.1)
probas = bayesClassifier.get_probas()
for dic in probas:
sums = 0.0
for item in dic:
sums += dic[item]
for item in dic:
dic[item] = dic[item] / sums
roc = ROC.Roc(Ytrue, probas, threshold, '')
roc.Roc_gen()
precision, recall, _ = precision_recall_curve(Ytrue, probas)
plt.step(recall, precision, color='b', alpha=0.2, where='post')
plt.fill_between(recall, precision, step='post', alpha=0.2, color='b')
plt.xlabel('Recall')
plt.ylabel('Precision')
plt.ylim([0.0, 1.05])
plt.xlim([0.0, 1.0])
plt.title('Precision Recall Curve')
return Ytrue, Ypred
def main():
path_boy = "F:\\study in school\\machine learning\\forstudent\\实验数据\\boynew.txt"
path_girl = "F:\\study in school\\machine learning\\forstudent\\实验数据\\girlnew.txt"
# height = []
# weight = []
# feetsize = []
x_boy = []
x_girl = []
label_boy = [] # 1表示男，0表示女
label_girl = []
readdata1(path_boy, x_boy, label_boy, 1)
readdata1(path_girl, x_girl, label_girl, 0)
x_boy = np.mat(x_boy)
x_girl = np.mat(x_girl)
m1 = x_boy.mean(0)
m0 = x_girl.mean(0)
S1 = (x_boy - m1[0]).T * (x_boy - m1[0])
S0 = (x_girl - m0[0]).T * (x_girl - m0[0])
Sw = S1 + S0
S_inverse = Sw.I
W = S_inverse * (m1 - m0).T
M1 = float(W.T * m1.T)
M0 = float(W.T * m0.T)
w_decision0 = (M0 + M1) / 2
path_boy_test = "F:\\study in school\\machine learning\\forstudent\\实验数据\\boy.txt"
path_girl_test = "F:\\study in school\\machine learning\\forstudent\\实验数据\\girl.txt"
x = []
label = []
readdata1(path_boy_test, x, label, 1)
readdata1(path_girl_test, x, label, 0)
label_test = []
y = x * W
errorcount = 0
for i in range(len(label)):
if float(y[i] > w_decision0):
label_test.append(1)
if label[i] != 1:
errorcount = errorcount + 1
else:
label_test.append(0)
if label[i] != 0:
errorcount = errorcount + 1
e_percentage = errorcount / len(label_test)
print('fisher测试集的错误率为%f' % e_percentage)
#留一法
loo = LeavePOut(p=1)
error = 0
for train, test in loo.split(x, label):
x_boy = []
x_girl = []
label_boy = [] # 1表示男，0表示女
label_girl = []
for i in train:
if label[i] == 1:
x_boy.append(x[i])
label_boy.append(1)
else:
x_girl.append(x[i])
label_girl.append(0)
x_boy = np.mat(x_boy)
x_girl = np.mat(x_girl)
m1 = x_boy.mean(0)
m0 = x_girl.mean(0)
S1 = (x_boy - m1[0]).T * (x_boy - m1[0])
S0 = (x_girl - m0[0]).T * (x_girl - m0[0])
Sw = S1 + S0
S_inverse = Sw.I
W = S_inverse * (m1 - m0).T
M1 = float(W.T * m1.T)
M0 = float(W.T * m0.T)
w_decision0 = (M0 + M1) / 2
for j in test:
if float(x[j] * W > w_decision0):
if label[j] != 1:
error = error + 1
else:
label_test.append(0)
if label[j] != 0:
error = error + 1
print('fisher留一法的错误率为%f' % (error / len(label)))
figure(3)
FPR, TPR = get_roc_fisher(W, w_decision0, x, label)
plot(FPR, TPR, label='fisher')
figure(5)
x1 = np.arange(130, 190, 0.01)
y1 = (w_decision0 - W[0] * x1) / W[1]
plot(x1, array(y1)[0])
plot(x1, x1 * float(W[1]) / float(W[0]))
for i in range(len(label)):
if label[i] == 1:
plot(float(x[i][0]), float(x[i][1]), 'o', color='r')
else:
plot(float(x[i][0]), float(x[i][1]), 'o', color='g')
a=(float(x[i][1])+float(x[i][0])*float(W[0])/float(W[1]))/\
(float(W[1])/float(W[0])+float(W[0])/float(W[1]))
b = a * float(W[1]) / float(W[0])
plot([float(x[i][0]), a], [float(x[i][1]), b], '--', color='0.75')
axis([140, 190, 35, 85])
Bayes()
class Test(object):
bayes = Bayes.Bayes()
def testingNB(self):
listOPosts, listClasses = self.bayes.loadDataSet()
myVocabList = self.bayes.createVocabList(listOPosts)
trainMat = []
for postinDoc in listOPosts:
trainMat.append(self.bayes.setOfWords2Vec(myVocabList, postinDoc))
p0V, p1V, pAb = self.bayes.trainNB0(array(trainMat),
array(listClasses))
testEntry = ['love', 'my', 'dalmation']
thisDoc = array(self.bayes.setOfWords2Vec(myVocabList, testEntry))
testEntry = ['stupid', 'garbage']
thisDoc = array(self.bayes.setOfWords2Vec(myVocabList, testEntry))
print testEntry, 'classified as: ', self.bayes.classifyNB(
thisDoc, p0V, p1V, pAb)
# 测试函数
def spamTest(self):
docList = []
classList = []
fullText = []
for i in range(1, 26):
# 导入文件 并解析成词列表
wordList = self.bayes.textParse(
open(Config.DATAS + 'NaiveBayes/email/spam/%d.txt' % i).read())
docList.append(wordList)
fullText.extend(wordList)
classList.append(1)
wordList = self.bayes.textParse(
open(Config.DATAS + 'NaiveBayes/email/ham/%d.txt' % i).read())
docList.append(wordList)
fullText.extend(wordList)
classList.append(0)
vocabList = self.bayes.createVocabList(docList) # create vocabulary
trainingSet = range(50)
testSet = [] # 随机构建测试函数
for i in range(10):
randIndex = int(random.uniform(0, len(trainingSet)))
testSet.append(trainingSet[randIndex])
del (trainingSet[randIndex])
trainMat = []
trainClasses = []
for docIndex in trainingSet: # 测试分类器
trainMat.append(
self.bayes.bagOfWords2VecMN(vocabList, docList[docIndex]))
trainClasses.append(classList[docIndex])
p0V, p1V, pSpam = self.bayes.trainNB0(array(trainMat),
array(trainClasses))
errorCount = 0
for docIndex in testSet: # classify the remaining items
wordVector = self.bayes.bagOfWords2VecMN(vocabList,
docList[docIndex])
if self.bayes.classifyNB(array(wordVector), p0V, p1V,
pSpam) != classList[docIndex]:
errorCount += 1
print "classification error", docList[docIndex]
e = float(errorCount) / len(testSet)
return e
print 'the error rate is: ', e
