# load the dataset
from sklearn import datasets
iris = datasets.load_iris()
data_X = iris.data
data_y = iris.target
np.unique(iris_y)
# initialize the Estimator
estimator = Estimator(param1=1, param2=2)
# get params of the existing estimator
estimator.estimated_param_
# train the estimator
estimator.fit(data_X_train, data_y_train)
# predict with estimator
estimator.predict(data_X_test)
# get estimator score - 1 is perfect prediction, 0 mean incorrect
estimator.score(data_X_test, data_y_test)
# The mean square error
np.mean((estimator.predict(data_X_test) - data_y_test) ** 2)
# get metrics for predictor
from sklearn import metrics
expected = digits.target[n_samples / 2:]
predicted = classifier.predict(data[n_samples / 2:])
print(metrics.classification_report(expected, predicted))
# Choose best alpha based on score
alphas = np.logspace(-4, -1, 6)
regr = linear_model.Lasso()
scores = [regr.set_params(alpha=alpha). \
fit(data_X_train, data_y_train). \
score(data_X_test, data_y_test) for alpha in alphas]
best_alpha = alphas[scores.index(max(scores))]
regr.alpha = best_alpha
regr.fit(data_X_train, data_y_train)
# to split data in train and test sets dynamically
# compute score method of an estimator from kfolds
# (total_number_of elements, numbers_of_folds)
kfold = cross_validation.KFold(len(X_digits), n_folds=3)
# kfold = [[train_indexes, test_indices], [train_indexes, test_indices], .. 3]
[svc.fit(X_digits[train], y_digits[train]).score(X_digits[test], y_digits[test])
for train, test in kfold]
# cross_vall_score use folds to validate data
cross_validation.cross_val_score(svc, X_digits, y_digits, cv=kfold, n_jobs=-1)
#### description of regularization parameter C
# Regularization is set by the C parameter: a small value for C means the
# margin is calculated using many or all of the observations around the
# separating line (more regularization);
# a large value for C means the margin is calculated on observations
# close to the separating line (less regularization).
# choose C value based on best score
# this is the same with the alpha above, but use 3 KFolds to find best param
from sklearn import cross_validation, svm
svc = svm.SVC(kernel='linear')
C_s = np.logspace(-10, 0, 10)
scores = list()
scores_std = list()
for C in C_s:
svc.C = C
# cv=3 - default for cross_val_score
this_scores = cross_validation.cross_val_score(svc, X, y, n_jobs=1)
scores.append(np.mean(this_scores))
scores_std.append(np.std(this_scores))
best_c = C_s[scores.index(max(scores))]
svc.C = best_c
svc.fit(data_X_train, data_y_train)
# SVMs can be used in regression # –SVR (Support Vector Regression)–,
# or in classification –SVC (Support Vector Classification).
#### different types of SVC kernels
# linear kernel
svc = svm.SVC(kernel='linear')
# polynomial kernel
scv = svm.SVC(kernel='poly', degree=3)
# RBF(Radial Basis Function) kernel
scv = svm.SVC(kernel='rbf')
# find the best param from the range with kfolds
from sklearn.grid_search import GridSearchCV
Cs = np.logspace(-6, -1, 10)
clf = GridSearchCV(estimator=svc, param_grid=dict(C=Cs), n_jobs=-1)
clf.fit(X_digits[:1000], y_digits[:1000])
clf.best_score_
clf.best_estimator_.C
# Prediction performance on test set is not as good as on train set
clf.score(X_digits[1000:], y_digits[1000:])
# use existing cross validation model. Find parameters dinamically based on
# estimator type. Note -CV at the end of estimator name
from sklearn import linear_model, datasets
lasso = linear_model.LassoCV()
dataset = datasets.load_diabetes()
X_data = dataset.data
y_data = dataset.target
lasso.fit(X_data, y_data)
# The estimator chose automatically its lambda:
lasso.alpha_
# clustering
from sklearn import cluster, datasets
iris = datasets.load_iris()
X_iris = iris.data
y_iris = iris.target
k_means = cluster.KMeans(n_clusters=3)
k_means.fit(X_iris)
print(k_means.labels_[::10])
print(y_iris[::10])