def do_windowing(window_size, begin, end):
gestures = PrepareDataSet.getlines()
lines = gestures.readlines()
count = 1
line = ''
return_xy = {}
for index_line in range(begin, end):
if count <= window_size:
if count == 1:
fields = lines[index_line].split(",")
if (end - index_line) >= window_size:
y.append(str(fields[32:]).replace("\\n", "").replace("\\r", "").replace("'", ""))
if count > 1:
line += "," + remove_labels_from_line(lines[index_line])
else:
line += remove_labels_from_line(lines[index_line])
count += 1
else:
count = 1
X.append(line.split(","))
line = ''
return_xy["X"] = copy.deepcopy(X)
return_xy["y"] = copy.deepcopy(y)
X[:] = []
y[:] = []
print "tamanho de x "+str(len(return_xy["X"]))
print "tamanho de y "+str(len(return_xy["y"]))
return return_xy
def execute():
gestures = PrepareDataSet.getlines()
lines = gestures.readlines()
instances_to_be_tested = []
for index_line in range(start_test_line_number, number_of_lines):
line = lines[index_line]
instances_to_be_tested.append(line)
prepared_instances_to_be_tested = prepare_instances_to_be_tested(instances_to_be_tested, False)
results = run_knn(prepared_instances_to_be_tested, False)
print "Percentual de acertos "\
+ str(get_percentage_corrects_predictions(results,
get_real_phase_of_gestures(instances_to_be_tested)))+"%"
from sklearn.metrics import confusion_matrix TRAINING = 0 TEST = 1 k = 5 # 3 files = [f for f in listdir(Pds.PATH) if isfile(join(Pds.PATH, f)) and "windowed" in f] # b1_va3_windowed windowed files.sort() print(files) # print(len(files)) data = [Pds.get_dataset(Pds.PATH+"/"+file) for file in files] data_x = [Pds.get_training_data(Pds.TRAINING_SLICE, data[index]['x'], type='x') for index in range(len(data))] print(data_x) data_y = [Pds.get_training_data(Pds.TRAINING_SLICE, data[index]['y'], type='y') for index in range(len(data))] clfs = [Knn.KNNClassifier(data_x[index][TRAINING], data_y[index][TRAINING], k) for index in range(len(data))] # clf = knn.KNNClassifier(data_x[1][1], data_y[1][1], k) # results = [] # for index in range(len(clfs)): # predictions = [] # for inner in range(len(data_x[index][1])): # predictions.append(clfs[index].classify(np.squeeze(np.asarray(data_x[index][TEST][inner]))))
x = [i for i in range(len(eigen_values))]
soma = 0
for index in range(15):
soma += var_exp[index]
print soma
plt.plot(x, y, linestyle='--', marker='o', color='b')
plt.ylabel("Porcentagem de Representacao")
plt.xlabel("Indice dos Autovalores")
plt.show()
# dataset = pds.get_dataset(pds.FILE)
dataset = pds.get_dataset("")
reduced_matrix = execute(dataset)
print ("final", reduced_matrix)
# with open(pds.PATH+"/a1_va3_reducedR.csv", 'w') as csvw:
# csvw = csv.writer(csvw, delimiter=',')
# csvw.writerows(reduced_matrix)
np.savetxt(pds.FILE_REDUCED, reduced_matrix, delimiter=',', fmt='%.8f')
print('y', dataset['y'])
outf = open(pds.FILE_REDUCED_PRED, 'w')
for index in range(len(dataset['y'])):
# -*- coding: utf-8 -*-
from sklearn import neighbors
import copy
from tools import PrepareDataSet
weight = 'distance'
n_neighbors = 7
percent_instances_to_trains_ = 80
window_size = 8
X = []
y = []
number_of_lines = PrepareDataSet.get_total_of_lines()
start_test_line_number = ((number_of_lines * percent_instances_to_trains_)/100)
print "teste começa nessa linha = " + str(start_test_line_number)
print "número total de linhas no arquivo = " + str(number_of_lines)
# lê todo o arquivo de dados do dataset setado no arquivo load_dataset no caso o a1_va3.csv
gestures = PrepareDataSet.getlines()
# pegar % do arquivo não janelado para treino
# refazer isso aqui está indo todo mundo pra o treinamento
def get_instances_to_train():
list_train = []
lines = gestures.readlines()
for index in range(1, start_test_line_number-1):
list_train.append(lines[index])
print "quantidade de linhas na memoria " + str(len(list_train))