def computerTrainError(self): posPath = '/Users/sunxinzi/Documents/Machine_Learning/Project/Sentiment Prediction/train/pos/' negPath = '/Users/sunxinzi/Documents/Machine_Learning/Project/Sentiment Prediction/train/neg/' testPath = '/Users/sunxinzi/Documents/Machine_Learning/Project/Sentiment Prediction/train/pos/' process = preprocess() posTrain = [] negTrain = [] testData = [] posTrain = process.getCleanTxt(posPath, 'pos') negTrain = process.getCleanTxt(negPath, 'neg') testData = process.getCleanTestData(testPath) classifier = naiveBayes() result = classifier.test(posTrain, negTrain, testData) errorCount = 0 for i in result: if i[1] == 'neg': errorCount = errorCount + 1 testPath = '/Users/sunxinzi/Documents/Machine_Learning/Project/Sentiment Prediction/train/neg/' testData = [] testData = process.getCleanTestData(testPath) result = classifier.test(posTrain, negTrain, testData) for i in result: if i[1] == 'pos': errorCount = errorCount + 1 return errorCount
def semTratamento(textos, polaridades, classes):
s1, s2 = naiveBayes(textos, polaridades, classes)
arq = open(pasta + 'resultados sem tratamento.txt', 'w')
arq.writelines(s1)
arq.write('\n')
arq.writelines(s2)
arq.close()
from modifyInput import modifier
from naiveBayes import naiveBayes
import pandas as pd
link = "Medical_data.csv"
testframe = pd.read_csv("test_medical.csv")
totalTestpoints = testframe.shape[0]
classifier = naiveBayes(link)
classifier.fit()
count = 0
for i in range(0, totalTestpoints):
original = testframe.iat[i, 0]
predicted = classifier.predict((testframe.iloc[i, 1:4]).tolist())
if predicted == original:
count = count + 1
print("accuracy=", 100 * count / totalTestpoints)
# generate 365 days worth of data
# this will take a format of [0, 1, 0, 1, 1, 0] - these are all the data variables
days = 365
while days > 0:
randList = lambda n: [randint(0, 1) for b in range(1, n+1)] # code attained at http://code.activestate.com/recipes/577944-random-binary-list/
oneDay = randList(20)
dailyMood.append(oneDay)
days -= 1
# let's assume there'll be 4 classes for our moods - angry, happy, sad, neutral
randMood = lambda n: [randint(0, 3) for b in range (1, n+1)]
target = randMood(365)
# test naive bayes on generated array
# instantiate object
clf = naiveBayes(dailyMood, target)
data_train, target_train, data_test, target_test = clf.split_train_test(dailyMood, target, 0.7) # use 70% of the data as training
# split class
# basically generates a dictionary where the key is a target class, and the values are the list of instances that fall in that class
splitClass = clf.classSplitter(data_train, target_train)
# get probabilities of each class appearing. This will probably be fumbled since we have a random set
classProbs = clf.classProbabilities(target_train)
condiProbs = clf.get_conditional_probs(splitClass)
# alright, now we have trained the set.
# let's test the effectiveness
accuracy = clf.test_errors(data_train, target_train, data_test, target_test)
print accuracy
# -*- coding: utf-8 -*-
"""
Created on Thu Jun 14 21:54:46 2018
@author: ssheh
Testimi i algoritmit te naiveBayes.py
"""
import csv
from naiveBayes import naiveBayes
reader = csv.DictReader(open('TestOrigins.csv'))
k = naiveBayes(shteguDataBaze="Origins.csv", AtributiKlase="Buys_Computer")
k.kalkulo_propabilitetin_AtributitKlase()
for row in reader:
k.hipoteza = row
k.Kalkulo_propabilitetin_kushtezues(k.hipoteza)
k.klasifiko()
print("\n\n")
from naiveBayes import naiveBayes
from knn import knn_string
from svm import runMachine
test_messages = [
"hey dude what's up",
"HEY BRO, WHATSUP! REPLY NOW AT 83147661764. WIN WIN WIN", "i Like pie",
"TEXT ME to win a FREE car free 4 lyfe yo 374237432"
]
# Demo Naive Bayes
print("Naive Bayes")
print("------------")
for msg in test_messages:
print(naiveBayes(msg))
print("K-Nearest Neighbors")
print("--------------------")
for msg in test_messages:
knn_string(msg)
print("Support Vector Machine")
print("----------------------")
for msg in test_messages:
runMachine(msg)
def bagOfWords(linesOfReal,linesOfFake, ngram, stopEnglish, stemed):
# create the transform / stop_words="english"
vectorizerReal = CountVectorizer(lowercase=True, stop_words=stopEnglish, analyzer='word',
ngram_range=(ngram, ngram), max_df=1.0, min_df=1,max_features=None)
trainReal = linesOfReal[:]
vectorizerReal.fit(trainReal) ##fit in vector
indexDictOfWordReal = vectorizerReal.vocabulary_ ##assign index for each word by vocab function
bag_of_words_real = vectorizerReal.transform(trainReal)
BoWOfReal = bag_of_words_real.toarray() ##bag of words array
uniqlistOfRealWords = vectorizerReal.get_feature_names() ## create unique list by feature function
#print("uniq real: ",len(uniqlistOfRealWords))
#print("uniq real: ",uniqlistOfRealWords)
# create the transform / stop_words="english"
vectorizerFake = CountVectorizer(lowercase=True, stop_words=stopEnglish, analyzer='word',
ngram_range=(ngram, ngram), max_df=1.0, min_df=1,max_features=None)
trainFake = linesOfFake[:]
vectorizerFake.fit(trainFake) ##fit in vector
indexDictOfWordFake = vectorizerFake.vocabulary_ ##assign index for each word by vocab function
bag_of_words_fake = vectorizerFake.transform(trainFake)
BoWOfFake = bag_of_words_fake.toarray() ##bag of words array
uniqlistOfFakeWords = vectorizerFake.get_feature_names() ## create unique list by feature function
countOfRealsDict = {}
countOfFakesDict = {}
frequenciesOfReal = np.sum(BoWOfReal,axis=0) ## sum counts of the same worlds
for word in indexDictOfWordReal.keys():
countOfRealsDict[word] = frequenciesOfReal[indexDictOfWordReal[word]]
frequenciesOfFake = np.sum(BoWOfFake,axis=0) ## sum counts of the same worlds
for word in indexDictOfWordFake.keys():
countOfFakesDict[word] = frequenciesOfFake[indexDictOfWordFake[word]]
testHeadLines = readCSV("test.csv", ngram, stemed)
#write results to the csv file
with open('output.csv', mode='w') as csv_file:
fieldnames = ['Id', 'Category']
writer = csv.DictWriter(csv_file, fieldnames=fieldnames)
writer.writeheader()
correctnessCount = 0
for line in testHeadLines.keys():
#stop_words="english"
vectorizerTest = CountVectorizer(lowercase=True, stop_words=stopEnglish, analyzer='word',
ngram_range=(ngram, ngram), max_df=1.0, min_df=1, max_features=None)
temp = []
temp.append(testHeadLines[line]["Id"])
#print(temp)
vectorizerTest.fit(temp) ##fit in vector
testindexDictOfWord = vectorizerTest.vocabulary_ ##assign index for each word by vocab function
test_bag_of_words = vectorizerTest.transform(temp)
test_BoW= test_bag_of_words.toarray() ##bag of words array
test_uniqlistOfWords = vectorizerTest.get_feature_names() ## create unique list by feature function
#print(testHeadLines[line]["Id"])
result = naiveBayes(countOfRealsDict, BoWOfReal, countOfFakesDict, BoWOfFake, testindexDictOfWord, test_BoW)
## write results to csv
writer.writerow({'Id': testHeadLines[line]["Id"], 'Category': result})
#print(result)
if result == testHeadLines[line]["Category"]:
correctnessCount += 1
#print("--------------------------------------")
uniqWordsofFiles = list(set(list(countOfRealsDict.keys()) + list(countOfFakesDict.keys())))##for bayes calculations
#print("uniq word Count: ", len(uniqWordsofFiles)) # feature names
print("correctness count: ", correctnessCount)
accuracy = calculationofAccuracy(correctnessCount, len(testHeadLines.keys()))
print("Accuracy = ", accuracy)
return [countOfRealsDict, countOfFakesDict, uniqlistOfRealWords, uniqlistOfFakeWords]
def test10Fold():
global allWords
splits = tenFoldCrossValidation()
count = 0
total = 0
print("Naive Bayes")
for split in splits:
nb = naiveBayes()
trainFeatures = []
trainClasses = []
testFeatures = []
testClasses = []
for example in split.train:
trainFeatures.append(example.features)
trainClasses.append(example.klass)
for example in split.test:
testFeatures.append(example.features)
testClasses.append(example.klass)
nb.train(trainFeatures, trainClasses)
nb.test(testFeatures, testClasses)
accuracy = nb.getCorrectCount() / len(testClasses)
total = total + accuracy
print("[INFO]\tFold ", str(count), " Accuracy:", str(accuracy))
count = count + 1
print("[INFO]\tAccuracy:", str(total / 10))
count = 0
total = 0
print("Random Forest")
for split in splits:
nb = RandomForest(100)
trainFeatures = []
trainClasses = []
testFeatures = []
testClasses = []
for example in split.train:
trainFeatures.append(example.features)
trainClasses.append(example.klass)
for example in split.test:
testFeatures.append(example.features)
testClasses.append(example.klass)
nb.train(trainFeatures, trainClasses)
nb.test(testFeatures, testClasses)
accuracy = nb.getCorrectCount() / len(testClasses)
total = total + accuracy
print("[INFO]\tFold ", str(count), " Accuracy:", str(accuracy))
count = count + 1
print("[INFO]\tAccuracy:", str(total / 10))
count = 0
total = 0
print("Neural 5")
for split in splits:
nb = neuralNetwork((5, ), 1000)
trainFeatures = []
trainClasses = []
testFeatures = []
testClasses = []
for example in split.train:
trainFeatures.append(example.features)
trainClasses.append(example.klass)
for example in split.test:
testFeatures.append(example.features)
testClasses.append(example.klass)
nb.train(trainFeatures, trainClasses)
nb.test(testFeatures, testClasses)
accuracy = nb.getCorrectCount() / len(testClasses)
total = total + accuracy
print("[INFO]\tFold ", str(count), " Accuracy:", str(accuracy))
count = count + 1
print("[INFO]\tAccuracy:", str(total / 10))
count = 0
total = 0
print("Neural 3")
for split in splits:
nb = neuralNetwork((3, ), 1000)
trainFeatures = []
trainClasses = []
testFeatures = []
testClasses = []
for example in split.train:
trainFeatures.append(example.features)
trainClasses.append(example.klass)
for example in split.test:
testFeatures.append(example.features)
testClasses.append(example.klass)
nb.train(trainFeatures, trainClasses)
nb.test(testFeatures, testClasses)
accuracy = nb.getCorrectCount() / len(testClasses)
total = total + accuracy
print("[INFO]\tFold ", str(count), " Accuracy:", str(accuracy))
count = count + 1
print("[INFO]\tAccuracy:", str(total / 10))
count = 0
total = 0
print("SVM")
for split in splits:
nb = svm()
trainFeatures = []
trainClasses = []
testFeatures = []
testClasses = []
for example in split.train:
trainFeatures.append(example.features)
trainClasses.append(example.klass)
for example in split.test:
testFeatures.append(example.features)
testClasses.append(example.klass)
nb.train(trainFeatures, trainClasses)
nb.test(testFeatures, testClasses)
accuracy = nb.getCorrectCount() / len(testClasses)
total = total + accuracy
print("[INFO]\tFold ", str(count), " Accuracy:", str(accuracy))
count = count + 1
print("[INFO]\tAccuracy:", str(total / 10))
def naiveBayesClassification(data):
# setup the data
# total length of dataset
totalLen = len(data)
# index of last element
lastElement = len(data[0]) - 1
# get all indices where row is not a spam (0)
zeroesIndices = np.where(data[:, lastElement] == 0)[0]
# get all indices where row is a spam (1)
onesIndices = np.where(data[:, lastElement] == 1)[0]
# store all non spam rows in a separate numpy array
zeroes = copy.deepcopy(data[zeroesIndices, :])
# store all spam rows in a separate numpy array
ones = copy.deepcopy(data[onesIndices, :])
# split train and test set 50% 50%
# with equal amount of 0s and 1s
# initialize train set
testSet = np.empty((0, len(data[0])), dtype=np.float64)
# stack half of 0s and 1s into the train set
testSet = np.vstack((testSet, zeroes[0:len(zeroes) // 2, :]))
testSet = np.vstack((testSet, ones[0:len(ones) // 2, :]))
# initialize test set
trainSet = np.empty((0, len(data[0])), dtype=np.float64)
# stack half of 0s and 1s into the test set
trainSet = np.vstack((trainSet, zeroes[len(zeroes) // 2:len(zeroes), :]))
trainSet = np.vstack((trainSet, ones[len(ones) // 2:len(ones), :]))
# extract labels for both sets
trainLabels = trainSet[:, -1]
testLabels = testSet[:, -1]
# # both sets ignore the class column
# trainSet= trainSet[:, :-1]
# testSet= testSet[:, :-1]
# total length of train and test set should
# be equal to total dataset length
assert (len(trainSet) + len(testSet) == len(data))
# uncomment to test 40% 60% distribution
# z = 0
# o = 0
# t = len(trainSet)
# for i in trainSet:
# print(i[len(i) - 1])
# if i[len(i) - 1] == 0:
# z+=1
# if i[len(i) - 1] == 1:
# o+=1
# print('z/t', z/t)
# print('o/t', o/t)
# return None
# get how many attributes there are in the given train set
numAttributes = len(trainSet[0]) - 1
# get the last column (class values) from the train set
lastColumn = trainSet[:, len(trainSet[0]) - 1]
# get only the unique values (eliminate duplicates)
classes = set(lastColumn)
# get how many classes there are for the given dataset
# and create a naive bayes class with that number of classes,
# classes, and number of attributes in the given set
NB = naiveBayes(len(classes), numAttributes, classes)
# train and predict using provided naive bayes code
NB.train(trainSet)
NB.trainingOutput()
predictions = NB.classify(testSet)
# compute confusion matrix and print it
confusionMatrix = confusion_matrix(testLabels, predictions)
print(confusionMatrix)
from naiveBayes import naiveBayes
from data import data
folds = 8
naiveBayesIris = naiveBayes()
print("Iris Dataset\n===================================\n")
# Get Iris data
irisData = data.read('./db/Iris/iris.csv')
# Train and test the model with Iris data
irisAccuracyScores = naiveBayesIris.crossval_predict(irisData["data"], irisData["labels"], folds)
# Calculate accuracy score
irisTotalAccuracy = 0.0
for i in irisAccuracyScores:
irisTotalAccuracy += i
irisTotalAccuracy = irisTotalAccuracy / len(irisAccuracyScores)
print("\n\nTotal accuracy: " +
str(round(irisTotalAccuracy * 100, 2)) + "%\n\n")
naiveBayesbanknote = naiveBayes()
print("Banknote Dataset\n===================================\n")
# Get Iris data
banknoteData = data.read('./db/banknote_authentication/banknote_authentication.csv')
# Train and test the model with banknote data
banknoteAccuracyScores = naiveBayesbanknote.crossval_predict(banknoteData["data"], banknoteData["labels"], folds)
# Calculate accuracy score
banknoteTotalAccuracy = 0.0
for i in banknoteAccuracyScores:
banknoteTotalAccuracy += i
banknoteTotalAccuracy = banknoteTotalAccuracy / len(banknoteAccuracyScores)
#plt.plot(cArr, supportVectors, label='C vs Support Vectors')
#plt.plot(epochs, devArr, label='Dev Error')
#plt.title('Error Rates for Unaveraged and Averaged Perceptron')
#plt.legend()
#plt.xlabel('C')
#plt.ylabel('Support Vectors')
#plt.show()
if algorithm == 2:
# Run on Python 3
knn_fit(npData, newTarget)
if algorithm == 3:
# Run on Python 3
m = gradient_booster(npData, newTarget)
if algorithm == 4:
# Run on Python 3
gnb = naiveBayes(npData, newTarget)
if algorithm == 5:
# Run on Python 3
kbest = kbestfeatures(npData, newTarget)
if algorithm == 6:
# Run on Python 2
model, prediction = logisticRegression(npData, newTarget)
# -*- coding: utf-8 -*-
"""
Created on Thu Jun 14 21:54:46 2018
@author: ssheh
Testimi i algoritmit te naiveBayes.py
"""
from naiveBayes import naiveBayes
k = naiveBayes(shteguDataBaze="tabele_shembull.csv", AtributiKlase="play")
k.kalkulo_propabilitetin_AtributitKlase()
k.hipoteza = {
"outlook": "sunny",
"temp": "high",
"humidity": "high",
"windy": "true"
}
k.Kalkulo_propabilitetin_kushtezues(k.hipoteza)
k.klasifiko()
import pandas as pd
from naiveBayes import naiveBayes
luis = pd.read_csv("haha.csv")
luis.dropna(axis=0, inplace=True)
luis.drop(['ACTIVITY'], axis=1, inplace=True)
data = luis.values.tolist()
target = [row[-1] for row in data]
for row in data:
del row[-1]
clf = naiveBayes(data, target)
data_train, target_train, data_test, target_test = clf.split_train_test(
data, target, 0.7) # use 70% of the data as training
# split class
# basically generates a dictionary where the key is a target class, and the values are the list of instances that fall in that class
splitClass = clf.classSplitter(data_train, target_train)
# get probabilities of each class appearing. This will probably be fumbled since we have a random set
classProbs = clf.classProbabilities(target_train)
condiProbs = clf.get_conditional_probs(splitClass)
# alright, now we have trained the set.
# let's test the effectiveness
accuracy = clf.test_errors(data_train, target_train, data_test, target_test)
print accuracy
programStart = time()
bowAll = bagOfWords(allSentencesList)
bowSpam = bagOfWords(spamSentencesList)
bowHam = bagOfWords(hamSentencesList)
# print(bowAll, bowSpam, bowHam, sep = "\n")
print()
count = 0
spamCount = 0
hamCount = 0
for testList in testSentencesList:
res = naiveBayes(preprocess(testList[1]), bowAll, bowSpam, bowHam)
# print(testList[0], res)
if (res == testList[0]):
if (((res == "ham") and (hamCount < 2))
or ((res == "spam") and (spamCount < 2))):
print("Input:", testList[1], end="")
print("Expected output:", testList[0])
print("Predicted output:", res)
print()
if (res == "ham"):
hamCount += 1
else:
spamCount += 1
count += 1
programEnd = time()
# -*- coding: utf-8 -*-
"""
Created on Thu Jun 14 21:54:46 2018
@author: ssheh
Testimi i algoritmit te naiveBayes.py
"""
import csv
from naiveBayes import naiveBayes
reader = csv.DictReader(open('TeDhenatTestuese.csv'))
#k = naiveBayes(shteguDataBaze = "Iris.csv", AtributiKlase = "Species" )
#k = naiveBayes(shteguDataBaze = "WaterBears.csv", AtributiKlase = "Species" )
k = naiveBayes(shteguDataBaze="TeDhenatTrajnuese.csv", AtributiKlase="Klasa")
k.kalkulo_propabilitetin_AtributitKlase()
#k.hipoteza = {"SepalLengthCm":"5.1","SepalWidthCm":"3.5","PetalLengthCm":"1.4","PetalWidthCm":"0.2"}
#k.hipoteza = {"SSI":"30.16","BTWa":"26.05","BTWs":"20.87","BTWp":"19.98","BTWL":"0.5","BTPA":"0.77"}
#k.hipoteza = {"Instanca1":"0","Instanca2":"1","Instanca3":"1","Instanca4":"2","Instanca5":"3","Instanca6":"5"}
for row in reader:
k.hipoteza = row
k.Kalkulo_propabilitetin_kushtezues(k.hipoteza)
k.klasifiko()
print("\n\n")
import sys
sys.path.append('/home/patcha/Dropbox/Doutorado/Codigos/Python/utils/')
import numpy as np
from dataManipulation import data
from naiveBayes import naiveBayes
# loading the data set
print 'Loading the dataset...'
irisAll = np.genfromtxt('/home/patcha/Datasets/Iris/iris.csv', delimiter=',')
iris = data(dataset=irisAll,
percTrain=0.7,
percVal=0,
percTest=0.3,
normType=None,
shuf=True,
posOut='last',
outBin=False)
print iris
nb = naiveBayes(iris.trainIn, iris.trainOut, iris.nClass)
splited = nb.splitByLabel()
stats = nb.statsByLabel(splited)
nb.getResult(stats, iris.testIn, iris.testOut)
# 学習データ（全体の90%）
Xtr = myData.X[:dtrNum]
Ytr = myData.Y[:dtrNum]
# 評価データ（全体の10%）
Xte = myData.X[dtrNum:]
Yte = myData.Y[dtrNum:]
#-------------------
#-------------------
# 3. ナイーブベイズの学習
# 事前確率の設定
priors = np.array([[0.5, 0.5]])
myModel = naiveBayes.naiveBayes(Xtr, Ytr, priors)
myModel.train()
#-------------------
#-------------------
# 4. ナイーブベイズの評価
print(f"学習データの正解率:{np.round(myModel.accuracy(Xtr,Ytr),decimals=2)}")
print(f"評価データの正解率:{np.round(myModel.accuracy(Xte,Yte),decimals=2)}")
#-------------------
#-------------------
# 5. 予測結果のCSVファイルへの出力
myModel.writeResult2CSV(
Xtr,
Ytr,
fName=f"../results/naiveBayes_result_train_{myData.dataType}.csv")
