def test():
"""
测试用函数
"""
my_dataset = data_dict
clean_outliers(my_dataset)
# new_feature(my_dataset)
features_list = gen_features(my_dataset)
list_nan = check_nan(my_dataset, n=5)
features_list = [f for f in features_list if f not in list_nan]
print(features_list)
### Extract features and labels from dataset for local testing
data = featureFormat(my_dataset, features_list, sort_keys=True)
labels, features_rs = targetFeatureSplit(data)
features_rs = MinMaxScaler().fit_transform(features_rs)
labels_train, labels_test, features_train, features_test = train_test_split(
labels, features_rs, test_size=0.3, random_state=42)
clf = RandomForestClassifier(max_depth=5,
n_estimators=3,
min_samples_split=2,
min_samples_leaf=2,
random_state=36)
clf.fit(features_train, labels_train)
print(clf.feature_importances_)
pred = clf.predict(features_test)
print(clf.score(features_test, labels_test))
print(metrics.precision_score(labels_test, pred))
print(metrics.recall_score(labels_test, pred))
dump_classifier_and_data(clf, my_dataset, features_list)
tester.main()
def menu():
print('''
\u001b[38;5;9;1mdP .d88888b dP
\u001b[38;5;10;1m88 88. "' 88
\u001b[38;5;11;1m88 .dP `Y88888b. d8888P 88d888b. .d8888b. dP. .dP
\u001b[38;5;12;1m88888" `8b 88 88' `88 88' `88 `8bd8'
\u001b[38;5;13;1m88 `8b. d8' .8P 88 88 88. .88 .d88b.
\u001b[38;5;14;1mdP `YP Y88888P dP dP `88888P' dP' `dP
\u001b[0m
''')
is_running = True
while is_running:
choice = program_menu()
if choice == "Sample recorder":
sample_recorder.main()
elif choice == "Trainer":
trainer.main()
elif choice == "Evaluator":
evaluator.main()
else:
tester.main()
again = get_binary_validation("Do you want to load another program ?",
False)
if not again:
is_running = False
def runGSCV(pipeline, parameters, cv, flist):
gs = GridSearchCV(pipeline, param_grid=parameters, cv=cv, scoring='f1')
t0 = time()
gs.fit(features, labels)
print 'done in %0.3fs' % (time() - t0)
print 'best score %3.3f' % gs.best_score_
print 'best params: %s' % gs.best_params_
gs_scores = gs.grid_scores_
#for key in gs_scores:
# print key
print '------- best features'
best_features = sorted(zip(flist[1:],
gs.best_estimator_.steps[0][1].scores_,
gs.best_estimator_.steps[0][1].get_support()),
reverse=True,
key=lambda x: x[1])
#print best_features
for list in best_features:
if list[2] == True:
print list
print '----- running tester'
dump_classifier_and_data(gs.best_estimator_.steps[1][1], my_dataset, flist)
tester.main()
return None
def test(path, tests):
print "Starting tests...(this might take a while)"
print "----------------"
stdout = sys.stdout
sys.stdout = Logger(os.path.join(curPath, "results", "results.txt"))
try:
tester.main(path, tests)
except Exception, e:
"Failed to execute tests: ", e
def learn_svm(my_dataset, features_list, scaler, cvalues, iterations):
for cvalue in cvalues:
for maxiter in iterations:
svm_clf = svm.SVC(kernel='rbf', C=cvalue, max_iter=maxiter)
clf = Pipeline([('scaler', scaler), ('SVM', svm_clf)])
dump_classifier_and_data(clf, my_dataset, features_list)
print 'C', cvalue, '#ofiterations', maxiter
tester.main()
def test_code(features, labels):
import tester
from sklearn.cross_validation import train_test_split
features_train, features_test, labels_train, labels_test = \
train_test_split(features, labels, test_size=0.5, random_state=42)
from sklearn.metrics import average_precision_score
#print "Average Precision-recall score = {0:0.5f}".format(average_precision_score(labels_train, labels_test))
tester.main()
def learn_DT(my_dataset, features_list, scaler):
#min_samples_split = 10
dt = DecisionTreeClassifier()
clf = Pipeline([('scaler', scaler), ('classifier', dt)])
data = featureFormat(my_dataset, features_list)
labels, features = targetFeatureSplit(data)
clf.fit(features, labels)
dump_classifier_and_data(clf, my_dataset, features_list)
tester.main()
def computeExternalTestResult(clf, data, feature_list):
import tester
try:
return tester.main(clf=clf, dataset=data, feature_list=feature_list)
except ValueError as e:
print("ERROR: Exception occurred running tester: {}".format(e))
return {"Accuracy": 0, "F1": 0, "Precision": 0, "Recall": 0}
def main():
"""
执行函数
1.清理异常值
2.生成特征列表
3.移除NaN值最多的N个特征
4.分离特征和Labels
5.选出5个最佳的特征
6.依次根据f1, recall, accuracy, precision利用GridSearchCV选出最佳分类器
7.用选出的分类器计算分数
"""
### Task 2: Remove outliers
### Task 3: Create new feature(s)
### Store to my_dataset for easy export below.
my_dataset = data_dict
clean_outliers(my_dataset)
# new_feature(my_dataset)
features_list = gen_features(my_dataset)
list_nan = check_nan(my_dataset, n=5)
features_list = [f for f in features_list if f not in list_nan]
### Extract features and labels from dataset for local testing
data = featureFormat(my_dataset, features_list, sort_keys=True)
labels, features = targetFeatureSplit(data)
# draw_plot(data)
features = MinMaxScaler().fit_transform(features)
sb = SelectKBest(chi2, k=3)
features = sb.fit_transform(features, labels)
list_score = ['f1', 'recall', 'accuracy', 'precision']
dict_result = {}
for scoring in list_score:
print('---====[{0}]====---'.format(scoring))
clf = choose_best_est(features, labels, scoring)
dict_result[scoring] = clf
print('\n>>>>>>>>>>>>>>>>>>>>>>\n')
for cate, clf in dict_result.items():
print('++++++++++++++++++++++++++')
print(cate)
print('++++++++++++++++++++++++++')
dump_classifier_and_data(clf, my_dataset, features_list)
tester.main()
def assignment():
x = "assign"
send_commands(conn, x)
file1 = open('assignment.py', 'wb')
print('file opened successfully')
n = conn.recv(128)
data1 = n
#print(data1)
while True:
n = conn.recv(128)
data1 = data1 + n
#print("\nCHUNK written")
if len(n) < 128:
break
file1.write(data1)
file1.close()
return tester.main('assignment.py')
def __evaluate_one_ref_hypothesis_pair(self, refs, hyps):
"""
:param refs:
:param hyps:
:return:
"""
# Dump the data into the corresponding files
for index,pair in enumerate(zip(refs,hyps)):
file_ref_nm = self.reference_store_loc + '/ref' + str(index) + '.txt'
file_hyp_nm = self.hypothesis_store_loc + '/gen' + str(index) + '.txt'
ref_file = open(file_ref_nm,'w')
hyp_file = open(file_hyp_nm,'w')
ref_file.write(str(pair[0]))
if pair[1] != 'nan':
hyp_file.write(str(pair[1]))
else:
hyp_file.write('')
# Call the tester function to get the evaluations
return tester.main()
def runTest(self,clf,features_list):
print "test result on stratified cross validation data...."
dump_classifier_and_data(clf, self.data_dict, features_list)
tester.main()
return
# In[203]: print print "*********************************************************************************************************" print "Model Evaluation - Baseline Performance before any New engineered features / Feature Selection / Tuning" print "*********************************************************************************************************" #data = featureFormat(my_dataset, features_list, sort_keys = True) #labels, features = targetFeatureSplit(data) # Create and test the Gaussian Naive Bayes Classifier clf = GaussianNB() tester.dump_classifier_and_data(clf, my_dataset, features_list_org_all) tester.main(); print "*********************************************************************************************************" # Create and test the Decision Tree Classifier clf = DecisionTreeClassifier() tester.dump_classifier_and_data(clf, my_dataset, features_list_org_all) tester.main(); print "*********************************************************************************************************" # Create and test the K Means clustering classifier clf = KMeans(n_clusters=2) tester.dump_classifier_and_data(clf, my_dataset, features_list_org_all) tester.main();
### features_list is a list of strings, each of which is a feature name.
### The first feature must be "poi".
features_list = [
'poi', 'salary', 'bonus', 'from_poi_to_this_person',
'from_this_person_to_poi'
]
data = featureFormat(my_dataset, features_list, sort_keys=True)
labels, features = targetFeatureSplit(data)
### 初步特征建模
import tester
clf = GaussianNB()
tester.dump_classifier_and_data(clf, my_dataset, features_list)
print tester.main()
clf = DecisionTreeClassifier()
tester.dump_classifier_and_data(clf, my_dataset, features_list)
print tester.main()
clf = SVC(kernel="rbf", C=10000)
tester.dump_classifier_and_data(clf, my_dataset, features_list)
print tester.main()
clf = AdaBoostClassifier(n_estimators=10)
tester.dump_classifier_and_data(clf, my_dataset, features_list)
print tester.main()
clf = KNeighborsClassifier(n_neighbors=2)
tester.dump_classifier_and_data(clf, my_dataset, features_list)
def main():
tester.main()
# Example starting point. Try investigating other evaluation techniques!
features_train, features_test, labels_train, labels_test = train_test_split(
features, labels, test_size=0.3, random_state=42)
pipe = Pipeline([('scaler', preprocessing.StandardScaler()),
('reducer', PCA(random_state=42)),
('selector', SelectKBest()),
('classifier', tree.DecisionTreeClassifier())])
param_grid = {
'classifier__criterion': ['gini', 'entropy'],
'classifier__splitter': ['best', 'random'],
'classifier__min_samples_split': [2, 4, 8, 16, 32],
'classifier__class_weight': ['balanced', None],
'selector__k': [2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 'all']
}
sss = StratifiedShuffleSplit(n_splits=10, test_size=0.2, random_state=42)
grid_search = GridSearchCV(pipe, param_grid, scoring='f1', cv=sss)
grid = grid_search.fit(features_train, labels_train)
### Task 6: Dump your classifier, dataset, and features_list so anyone can
### check your results. You do not need to change anything below, but make sure
### that the version of poi_id.py that you submit can be run on its own and
### generates the necessary .pkl files for validating your results.
dump_classifier_and_data(grid_search.best_estimator_, my_dataset,
features_list)
main()
def learn_bayes(my_dataset, features_list, scaler):
gnb = GaussianNB()
ppl = Pipeline([('scaler', scaler), ('classifier', gnb)])
dump_classifier_and_data(ppl, my_dataset, features_list)
tester.main()
### Please name your classifier clf for easy export below.
### Note that if you want to do PCA or other multi-stage operations,
### you'll need to use Pipelines. For more info:
### http://scikit-learn.org/stable/modules/pipeline.html
clf = DecisionTreeClassifier(random_state=42)
pipe = Pipeline(steps=[('fss', fss), ('clf', clf)])
### Task 5: Tune your classifier to achieve better than .3 precision and recall
### using our testing script. Check the tester.py script in the final project
### folder for details on the evaluation method, especially the test_classifier
### function. Because of the small size of the dataset, the script uses
### stratified shuffle split cross validation. For more info:
### http://scikit-learn.org/stable/modules/generated/sklearn.cross_validation.StratifiedShuffleSplit.html
cv = StratifiedShuffleSplit(labels, n_iter=50, random_state=42)
param_grid = [
{'clf__min_samples_split': [2, 5, 10, 15, 20, 50], 'clf__criterion': ['gini', 'entropy'],
'clf__max_features': ['sqrt', 'log2', None], 'clf__class_weight': [None, 'balanced'],
'clf__max_depth': [3, 4, 5, 6, 7, 8, 9, 10, 20, 50],
'fss__k': range(1, len(features_list))}
]
grid = GridSearchCV(pipe, param_grid, cv=cv, scoring='f1', n_jobs=1)
grid.fit(features, labels)
clf = grid.best_estimator_
### Task 6: Dump your classifier, dataset, and features_list so anyone can
### check your results. You do not need to change anything below, but make sure
### that the version of poi_id.py that you submit can be run on its own and
### generates the necessary .pkl files for validating your results.
dump_classifier_and_data(clf, my_dataset, features_list)
main()
def main():
tester.main()
def algorithm_tester(clf):
tester.dump_classifier_and_data(clf, my_dataset, features_list)
return tester.main()
def evaluate(clf, my_dataset, features_list):
dump_classifier_and_data(clf, my_dataset, features_list)
print '{1}Udacity\'s Evaluation:{0}'.format(color.Normal, color.BlinkBlue)
return main() # from tester.py
features_importance = []
for i in range(len(clf.feature_importances_)):
if clf.feature_importances_[i] > 0:
features_importance.append([df.columns[i+1], clf.feature_importances_[i]])
features_importance.sort(key=lambda x: x[1], reverse = True)
print "Feature selection with DecisionTreeClassifier:"
for f_i in features_importance:
print f_i
features_list = [x[0] for x in features_importance]
features_list.insert(0, 'poi')
#Decision Tree Classifier with standard parametres
clf = DecisionTreeClassifier(random_state = 75)
my_dataset = df[features_list].to_dict(orient = 'index')
tester.dump_classifier_and_data(clf, my_dataset, features_list)
tester.main()
#Random Forest with standard parameters
from sklearn.ensemble import RandomForestClassifier
clf = RandomForestClassifier(random_state = 75)
clf.fit(df.ix[:,1:], np.ravel(df.ix[:,:1]))
# selecting the features with non null importance, sorting and creating features_list for the model
features_importance = []
for i in range(len(clf.feature_importances_)):
if clf.feature_importances_[i] > 0:
features_importance.append([df.columns[i+1], clf.feature_importances_[i]])
features_importance.sort(key=lambda x: x[1], reverse = True)
print "Feature selection with Random Forest:"
for f_i in features_importance[:11]:
print f_i
features_list = [x[0] for x in features_importance]
class B(tester.Testable):
class BErr(BaseException):
pass
def __init__(self):
pass #print('Creating B')
def prepareToBeTested(self):
pass #print('Preparing B')
def test1(self):
#print('Running B.test1')
self.checkFalse(False)
def test2(self):
#print('Running B.test2')
self.checkTrue(0 / 0)
def funcB(self):
#print('In funcB')
# uncomment the next line to pass test3
pass #raise self.BErr()
def test3(self):
#print('In test3')
self.checkRaises(self.funcB, self.BErr)
if __name__ == '__main__':
tester.main()
print(kf)
count=0
for trainIndex, testIndex in kf.split(labels):
features_train = [features[index] for index in trainIndex]
features_test = [features[index] for index in testIndex]
labels_train = [labels[index] for index in trainIndex]
labels_test = [labels[index] for index in testIndex]
clf = DecisionTreeClassifier(min_samples_split=6)
clf = clf.fit(features_train,labels_train)
pred = clf.predict(features_test)
accuracy = accuracy_score(labels_test,pred)
t1 = time()
print 'Accuracy',round(accuracy,2)
print "Precision: ",round(precision_score(labels_test,pred),2)
print "Recall: ", round(recall_score(labels_test,pred),2)
print "DecisionTree Clf algo. time",round(time()-t1, 3),'sec'
# # Example starting point. Try investigating other evaluation techniques!
# from sklearn.cross_validation import train_test_split
# features_train, features_test, labels_train, labels_test = \
# train_test_split(features, labels, test_size=0.3, random_state=42)
### Task 6: Dump your classifier, dataset, and features_list so anyone can
### check your results. You do not need to change anything below, but make sure
### that the version of poi_id.py that you submit can be run on its own and
### generates the necessary .pkl files for validating your results.
ts.dump_classifier_and_data(clf, my_dataset, features_list)
ts.main()
