def predict(self, X: Union[DataFrame, Series]) -> Union[Series, Any]:
"""
Predict y value for a set of instances
:param X: set of instances which y value wants to be predicted (use pandas Series for individual instances)
:return: returns the list of predicted values for each instance in X (if X is just an individual instance it just
returns its predicted value)
"""
try:
getattr(self, "X_")
(X_train_aux, target_attribute) = _generate_train_set(self.X_, self.y_, self.target_attribute)
# noinspection PyPep8Naming
X_train: DataSet = DataSet()
X_train.load_from_pandas_df(X_train_aux, self.attribute_info, self.attribute_list)
except AttributeError:
raise RuntimeError("You must train classifier before predicting data!")
# X is a set
if X.__class__ == DataFrame:
y = X.apply(lambda row: naive_bayes_classifier(X_train, row, self.target_attribute), axis=1)
# X is an instance
else:
y = naive_bayes_classifier(X_train, X, self.target_attribute)
return y
def cross_validation_n_1(setbuilt, n): allset = setbuilt[0] classlist = setbuilt[1] vasize = len(allset) // n; for i in range(0, len(allset), vasize): validset = allset[i:i+vasize-1] trainset = [sam for sam in allset if sam not in validset] nbcc = nbc.naive_bayes_classifier(classlist, trainset, validset) nbcc.train() print 'rate of %d patch: %f' % (i // vasize, nbcc.validate())
def cross_validation_n_1(setbuilt, n):
allset = setbuilt[0]
classlist = setbuilt[1]
vasize = len(allset) // n
for i in range(0, len(allset), vasize):
validset = allset[i:i + vasize - 1]
trainset = [sam for sam in allset if sam not in validset]
nbcc = nbc.naive_bayes_classifier(classlist, trainset, validset)
nbcc.train()
print 'rate of %d patch: %f' % (i // vasize, nbcc.validate())
def fitness(init_population, population_size, num_features, dataset):
best_accuracy = 0
best_solution = [0 for i in range(num_features)]
fitness_list = []
for row in init_population:
accuracy = nb.naive_bayes_classifier(dataset, num_features, row)
if accuracy > best_accuracy:
best_accuracy = accuracy
best_solution = row
fitness_list.append(accuracy)
return fitness_list, best_accuracy, best_solution
import naive_bayes_classifier as nb
(script_name,model_file,test_file)=argv
prior_word_conditional_prob_hash=pickle.load(open(model_file,"rb"))
feature_rank_hash=pickle.load(open("feature_ranks.p","rb"))
prior_probabilities=pickle.load(open("prior_class_probabilities.p","rb"))
count=-1
with open(test_file, 'rb') as csvfile:
review_reader = csv.reader(csvfile, delimiter='\t', quotechar='"')
#review_reader.next()
for row in review_reader:
count=count+1
if count==0:
print "Id,Category"
continue
#rows=row[0].split(",",1)
review_content=""
#if len(rows)==2:
# review_content=rows[1]
#elif len(rows)==1:
# review_content=rows[0]
review_content=row[0]
words=preprocessor.pre_process_sentence(review_content)
negative_words=pickle.load(open("negative_words.p","rb"))
positive_words=pickle.load(open("positive_words.p","rb"))
class_data=nb.naive_bayes_classifier(prior_probabilities,prior_word_conditional_prob_hash,words,positive_words,negative_words )
#print str(count)+","+class_data.keys()[0]
print class_data.keys()[0]+","+review_content
#break
for i in range(0, len(y_train)):
# y_train[i] is the index of the class_vector element we need to append
class_vector_train.append(class_vector[int(y_train[i])])
for i in range(0, len(y_test)):
# y_test[i] is the index of the class_vector element we need to append
class_vector_test.append(class_vector[int(y_test[i])])
print('class_vector_train: ', class_vector_train)
print('class_vector_test: ', class_vector_test)
print('Proportion of dataset selected to be included in test dataset: ', test_size)
# Bayes classifier
class_label_vector_bayes = bayes_classifier(X_train, class_vector_train, X_test)
# Naive Bayes classifier
class_label_vector_naive = naive_bayes_classifier(X_train, class_vector_train, X_test)
# Computation of accuracies of each method
accuracy_Bayes = compare_class_vectors(class_label_vector_bayes, class_vector_test)
accuracy_Naive_Bayes = compare_class_vectors(class_label_vector_naive, class_vector_test)
print('class_vector_test: ', class_vector_test)
print('class_label_vector Bayes: ', class_label_vector_bayes)
print('class_label_vector Naive Bayes: ', class_label_vector_naive)
print('Accuracy of Bayes is: ', accuracy_Bayes)
print('Accuracy of Naive Bayes is: ', accuracy_Naive_Bayes)
import pandas as p
from naive_bayes_classifier import naive_bayes_classifier
train = p.read_csv("trainBruno.csv")
test = p.read_csv("testBruno.csv")
#print(train)
machine = naive_bayes_classifier()
machine.train(train)
#machine.predict(1,1)
predict = machine.test(test)
#predict = machine.crossvalidation()
print(predict)
index = random.randint(0, len_chromosome - 1)
chromosome[index] = (chromosome[index] + 1) % 2
def geneticAlgo(population_size, num_features, dataset):
# First ---> Evaluate fitness
for i in range(NUM_ITERATIONS):
init_population = [[random.randint(0, 1) for i in range(num_features)]
for j in range(population_size)]
fitness_list, best_accuracy, best_solution = fitness(
init_population, population_size, num_features, dataset)
selected_indices = selection(fitness_list, dataset, population_size)
new_population = [init_population[i][:] for i in selected_indices]
crossover(new_population)
mutation(new_population)
return best_accuracy, best_solution
if __name__ == '__main__':
print("Enter name of dataset: ")
dataset_name = input()
dataset = pandas.read_csv(dataset_name)
num_features = len(dataset.columns) - 1
unmodified = [1 for i in range(num_features)
] # Since the last column is the output column
print("Naive Bayes mean accuracy without feature selection: ",
nb.naive_bayes_classifier(dataset, num_features, unmodified))
print("Naive Bayes mean accuracy with feature selection: ",
geneticAlgo(30, num_features, dataset))
population_size = 30