##
## As always, start with the documentation and try to reproduce the
## results there:
##
## http://scikit-learn.org/stable/modules/svm.html
##

## Goal: train and tune an SVM on a new data set.
from sklearn.svm import SVC

## 1. download the data (predict Leukemia from a set of gene expression data):
from sklearn.datasets import fetch_mldata
leuk = fetch_mldata('leukemia', data_home='./')

## check a basic classifier - just to have an idea of its performance

clf = SVC(...put your favourite parameters here...)
clf.fit(leuk.data, leuk.target)

## 2. We want to find the best parameters for our classifier. Let's
## try "grid search": for a number of combinations of parameters, we
## estimate the performance and choose the best combination.

from sklearn.cross_validation import train_test_split
from sklearn.grid_search import GridSearchCV
from sklearn.metrics import classification_report

X_train, X_test, y_train, y_test = train_test_split(leuk.data, leuk.target, test_size=0.25, random_state=0)

# Set the parameters by cross-validation
tuned_parameters = [{'kernel': ['rbf'], 'gamma': [1e-3, 1e-4],
                     'C': [1, 10, 100, 1000]},
                    {'kernel': ['linear'], 'C': [1, 10, 100, 1000]}]
scores = ['accuracy', 'roc_auc']

for score in scores:
    print("# Tuning hyper-parameters for %s" % score)
    print()

    clf = GridSearchCV(SVC(C=1), tuned_parameters, cv=3, scoring=score)
    clf.fit(X_train, y_train)

    print("Best parameters set found on development set:")
    print()
    print(clf.best_estimator_)
    print()
    print("Grid scores on development set:")
    print()
    for params, mean_score, scores in clf.grid_scores_:
        print("%0.3f (+/-%0.03f) for %r" % (mean_score, scores.std() / 2, params))
    print()

    print("Detailed classification report:")
    print()
    print("The model is trained on the full development set.")
    print("The scores are computed on the full evaluation set.")
    print()
    y_true, y_pred = y_test, clf.predict(X_test)
    print(classification_report(y_true, y_pred))
    print()

##
## TODO: try another approach to parameter tuning:
## check the random search method (3.2.2 at http://scikit-learn.org/stable/modules/grid_search.html )
## and use it!