def print_pca_variance():
# reading the data from sql table using the get_all_data method
X_train, y_train, train_le, X_test, y_test, test_le = data_accessor_util.get_all_data_sets()
classes = list(test_le.classes_)
# Converting data to numpy
(X_train, y_train, X_test, y_test) = data_accessor_util.convert_data_sets_to_numpy(X_train, y_train, X_test, y_test)
n_features = X_test.shape[1]
# this block need to be used once to create the plots
pca = PCA(n_components=n_features)
pca.fit(X_train)
var = pca.explained_variance_ratio_
cumulative_var = np.cumsum(var)
# To enable creating figures on the server
matplotlib.use('Agg')
fig = plt.figure(1)
plt.plot(var)
plt.title('individual scree plot')
plt.xlabel('principal components')
plt.ylabel('proportion of variance explained')
plt.savefig("proportion_variance.jpg")
fig2 = plt.figure(2)
plt.plot(cumulative_var)
plt.title('commulative scree plot')
plt.xlabel('principal components')
plt.ylabel('commulative proportion of variance explained')
plt.savefig("proportion_Variance_com.jpg")
def read_data_perform_pca(var_percentage = 0.95): # reading the data from sql table using the get_all_data method X_train, y_train, train_le, X_test, y_test, test_le = data_accessor_util.get_all_data_sets() classes = list(test_le.classes_) # Converting data to numpy (X_train, y_train, X_test, y_test) = data_accessor_util.convert_data_sets_to_numpy(X_train, y_train, X_test, y_test) n_features = X_test.shape[1] # this block need to be used once to create the plots pca = PCA(n_components=n_features) pca.fit(X_train) var = pca.explained_variance_ratio_ cumulative_var = np.cumsum(var) # arg max returns index, add 1 b/c indices start at 0 percentage_retained = var_percentage N_reduced = np.argmax(cumulative_var>percentage_retained)+1; pca = PCA(n_components=N_reduced) X_train = pca.fit_transform(X_train) X_test = pca.transform(X_test) return X_train, y_train, train_le, X_test, y_test, test_le
from sklearn.model_selection import GridSearchCV
from sklearn.decomposition import PCA
from sklearn.metrics import classification_report
# Internal
sys.path.append(os.path.realpath("%s/.." % os.path.dirname(__file__)))
from util import data_accessor_util
#-------------------------
# Globals
#-------------------------
# Get data
(train_X, train_Y, train_le, test_X, test_Y,
test_le) = data_accessor_util.get_all_data_sets()
classes = list(test_le.classes_)
print test_le.inverse_transform([0, 1, 2, 3, 4, 5])
# Convert to numpy
(train_X, train_Y, test_X,
test_Y) = data_accessor_util.convert_data_sets_to_numpy(
train_X, train_Y, test_X, test_Y)
Cs = {"C": np.arange(10**-5, 10**-1, 0.005)}
# PCA
print "PCA"
pca = PCA(n_components=train_X.shape[1])
train_X = pca.fit_transform(train_X)
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.linear_model import LogisticRegression
from sklearn.gaussian_process import GaussianProcessClassifier
from matplotlib.colors import ListedColormap
from sklearn.datasets import make_moons, make_circles, make_classification
from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis
from sklearn.metrics.classification import accuracy_score, log_loss
from sklearn.gaussian_process.kernels import RBF
DEBUG = 1
# reading the data from sql table using the get_all_data method
X_train, y_train, train_le, X_test, y_test, test_le = data_accessor_util.get_all_data_sets()
# X = df.drop(['genre'], axis=1)
# y = df['genre']
# le = preprocessing.LabelEncoder()
# y = le.fit_transform(y)
X_train = X_train.drop('songID', axis=1)
X_test = X_test.drop('songID', axis=1)
h = .02 # step size in the mesh
# list of methods used, they correspond to the classifiers list
methods = ["Logistic Regression","Decision Tree", "Random Forest",
"Linear SVM", "RBF SVM", "Neural Net, MLP",
"Gaussian Process", "Gaussian Naive Bayes", "QDA","AdaBoost"]
# add these two methods as well
# Xgboost
# Gradient boosting