#LOADING THE DATASET
X=sio.loadmat('/home/set.mat')['x']
s_y=sio.loadmat('/home/set.mat')['y']
y=np.ravel(s_y).astype(int)


#PIPELINE
clf = Pipeline([('rcl', RobustScaler()),
                ('clf', RandomForestClassifier(random_state=0, n_jobs=-1))])


#OPTIMIZATION
sss_outer = StratifiedShuffleSplit(n_splits=10, test_size=0.1, random_state=0)
sss_inner = StratifiedShuffleSplit(n_splits=10, test_size=0.1, random_state=0)


parameters = {'clf__n_estimators': sp_randint(170, 300),
              'clf__max_features': sp_randint(5, 20),
                 }
n_iter_search = 30


inner_rs = RandomizedSearchCV(clf, param_distributions=parameters, n_iter=n_iter_search, cv=sss_inner, random_state=0, n_jobs=-1)
inner_rs.fit(X,y)
outer_rs = cross_val_score(inner_rs, X, y, cv=sss_outer)

print(inner_rs.best_params_)
print(outer_rs.mean())