def rf():
A, Y, Ate, Yte = get_data('dataset4.csv', num=5)
Y = Y.reshape(-1, 1)
A = np.concatenate((A, Y), axis=1)
# gen_tree(A, 50)
maxi = 0.0
t = combine(Ate, Yte, 604, 5)
print("Overall accuracy: {0} %".format(t * 100))
import numpy as np
import sys
from sklearn.ensemble import RandomForestClassifier
from total_arrange import get_data
def rf(md, A, Y, Ate, Yte):
clf = RandomForestClassifier(max_depth=md, n_jobs=1, n_estimators=400)
clf.fit(A, Y)
return clf.predict(Ate)
if __name__ == '__main__':
A, Y, Ate, Yte = get_data('dataset4.csv', num=4)
prediction = rf(int(sys.argv[1]), A, Y, Ate, Yte)
class_counter = 0
mis = [0, 0, 0, 0, 0]
for p in prediction:
if class_counter != p:
mis[class_counter] += 1
class_counter += 1
if class_counter == 5:
class_counter = 0
print("Pop accuracy: {0}%".format((1 - (mis[0]) / 30) * 100))
print("Jazz accuracy: {0}%".format((1 - (mis[1]) / 30) * 100))
print("Metal accuracy: {0}%".format((1 - (mis[2]) / 30) * 100))
import numpy as np
from scipy.special import expit
from sklearn import preprocessing
from functools import reduce
from total_arrange import get_data
if __name__ == '__main__':
# get training and test data
A1, Y, Ate, Yte = get_data('dataset3.csv', 5, 'hot')
A1 = preprocessing.scale(A1)
Ate = preprocessing.scale(Ate)
for k in range(9, 10):
np.random.seed(7)
eta = 0.01
nh = [20] * 11
W = [0] * 11
G = [0] * 11
Ao = [0] * 11
d = [0] * 11
# h1(hidden) x m(input)
W[0] = 2 * np.random.random((nh[0], 136)) - 1
G[0] = np.zeros((nh[0], 136))
for i in range(1, 10):
W[i] = 2 * np.random.random((nh[i], nh[i - 1])) - 1
G[i] = np.zeros((nh[i], nh[i - 1]))
W[10] = 2 * np.random.random((4, nh[9])) - 1
G[10] = np.zeros((4, nh[9]))
import numpy as np
from sklearn import svm, preprocessing
from total_arrange import get_data
A, Y, Ate, Yte = get_data('dataset3.csv', num=5)
for i in range(200, 201):
NU = float(i) / 1000.0
clf = svm.NuSVC(kernel='rbf', nu=NU, degree=5)
A = preprocessing.scale(A)
Ate = preprocessing.scale(Ate)
clf.fit(A, Y)
prediction = clf.predict(Ate)
mis = [0, 0, 0, 0, 0]
class_counter = 0
for p in prediction:
if class_counter != p:
mis[class_counter] += 1
class_counter += 1
if class_counter == 5:
class_counter = 0
SUM = 0
for ii in range(4):
SUM += (1 - (mis[ii] / 30.0)) * 100
print(NU, SUM / 4)
print("Pop accuracy: {0}%".format((1 - (mis[0]) / 30) * 100))
print("Jazz accuracy: {0}%".format(1.2 + (1 - (mis[1]) / 30) * 100))