def main(argv):
args = parser.parse_args(argv[1:])
batch_size = args.batch_size
(train_x, train_y), (test_x, test_y), unlabeled = meo_data.load_data()
# construct classifier
classifier = tf.estimator.Estimator(model_fn=model.story_model,
params={
'feature_columns':
fc.story_model_columns(
train_x, test_x, unlabeled),
'n_classes':
args.classes,
})
# train model
classifier.train(
input_fn=lambda: meo_data.train_input_fn(train_x, train_y, batch_size),
steps=args.train_steps)
# evaluate and print results
eval_result = classifier.evaluate(
input_fn=lambda: meo_data.eval_input_fn(test_x, test_y, batch_size))
m.print_eval(eval_result)
# predict classes in test data, print a random sample
p.predict(classifier, test_x, test_y, unlabeled, batch_size)
def main(argv):
args = parser.parse_args(argv[1:])
batch_size = args.batch_size
(train_x, train_y), (test_x, test_y), unlabeled = meo_data.load_data()
# construct classifier
classifier = tf.estimator.Estimator(
model_fn=model.story_model,
params={
'feature_columns': fc.story_model_columns(train_x, test_x,
unlabeled),
'n_classes': args.classes,
})
# train model
classifier.train(
input_fn=lambda: meo_data.train_input_fn(train_x, train_y, batch_size),
steps=args.train_steps)
# evaluate and print results
eval_result = classifier.evaluate(
input_fn=lambda: meo_data.eval_input_fn(test_x, test_y, batch_size))
m.print_eval(eval_result)
# predict classes in test data, print a random sample
p.predict(classifier, test_x, test_y, unlabeled, batch_size)
def predict(species, data_name):
'''
This function makes use of the predictions file to predict the
reliability of a given set of observations using a pre trained model.
:param species:
:param data_name:
:return:
'''
p.predict(species, data_name)
cwd = os.getcwd()
os.remove(os.path.join(cwd, 'temp.csv'))
def upload_file():
if request.method == 'POST':
# check if the post request has the file part
if 'file' not in request.files:
flash('No file part')
return redirect(request.url)
file = request.files['file']
# if user does not select file, browser also
# submit an empty part without filename
if file.filename == '':
flash('No selected file')
return redirect(request.url)
if file and allowed_file(file.filename):
filename = secure_filename(file.filename)
file.save(os.path.join(app.config['UPLOAD_FOLDER'], filename))
predicts = predictions.predict(
os.path.join(app.config['UPLOAD_FOLDER'], filename))
facts = get_poke_facts.get_facts(predicts[0])
print(facts)
return render_template("pokemon.html",
facts=facts,
pokemon_name=predicts[0],
image_data=predicts[1],
pokemon_percentage=predicts[2],
image_pokemon=predicts[3])
#return 'Jeg tror det er en: {}'.format(str(predicts[0]))
return render_template("upload.html", )
def predict():
messages = request.get_json(force=True)
messages = delete_empty_messages(messages)
keys = list(messages.keys())
values = [messages[key] for key in keys]
if values:
toxicity = predictions.predict(values)
toxic_messages = {
keys[index]: float(toxic)
for index, toxic in enumerate(toxicity)
}
return jsonify(toxic_messages)
else:
return jsonify([])
def query():
if request.method == 'GET':
error = 0
error_msg = ''
q = request.args.get('q')
if q:
result = ip_info.find_one({'ip': q})
if result:
results = {
'ip': result['ip'],
'country': result['country'],
'city': result['city'],
'country_id': result['country_id']
}
else:
result = otx.get_indicator_details_full(IndicatorTypes.IPv4, q)
if result:
results = {
'ip': result['general']['indicator'],
'country': result['general']['country_name'],
'city': result['general']['city']
}
ip.search_ip(q)
else:
results = {}
error = 1
if error:
city_pre = ''
else:
city_pre = predictions.predict(results['city'])
city_pre = city_pre[0:2]
result = []
for cp in city_pre:
r = ip_info.find({'city': cp[0]})
for rs in r[0:10]:
result.append(rs)
return render_template('index.html',
q=q,
error=error,
error_msg=error_msg,
results=results,
pre=result,
city_pre=city_pre)
else:
return render_template('index.html')
def post(self, request):
form = PortfolioForm(request.POST)
if form.is_valid():
holding = form.cleaned_data['holding']
if holding != 0:
span = 'Short-Term'
else:
span = 'Long-Term'
#asset1
asset1 = form.cleaned_data['asset1']
name1 = dict(form.fields['asset1'].widget.choices)[asset1]
weight1 = form.cleaned_data['weight1']
# asset2
asset2 = form.cleaned_data['asset2']
name2 = dict(form.fields['asset2'].widget.choices)[asset2]
weight2 = form.cleaned_data['weight2']
# asset3
asset3 = form.cleaned_data['asset3']
name3 = dict(form.fields['asset3'].widget.choices)[asset3]
weight3 = form.cleaned_data['weight3']
# asset4
asset4 = form.cleaned_data['asset4']
name4 = dict(form.fields['asset4'].widget.choices)[asset4]
weight4 = form.cleaned_data['weight4']
# Short-term holding
if span == 'Short-Term':
start_date = datetime.today() + timedelta(days=-100)
end_date = datetime.today()
#holidays = get_calendar('USFederalHolidayCalendar').holidays(start_date, end_date)
#period = [x for x in period if x not in holidays.date]
#period = pd.DatetimeIndex(period)
prices1 = web.DataReader(asset1, 'yahoo', start_date, end_date)
df1 = pd.DataFrame(predict(prices1, holding))
df1.rename(columns={df1.columns[0]: 'Close1'}, inplace=True)
prices2 = web.DataReader(asset2, 'yahoo', start_date, end_date)
df2 = pd.DataFrame(predict(prices2, holding))
df2.rename(columns={df2.columns[0]: 'Close2'}, inplace=True)
prices3 = web.DataReader(asset3, 'yahoo', start_date, end_date)
df3 = pd.DataFrame(predict(prices3, holding))
df3.rename(columns={df3.columns[0]: 'Close3'}, inplace=True)
prices4 = web.DataReader(asset4, 'yahoo', start_date, end_date)
df4 = pd.DataFrame(predict(prices4, holding))
df4.rename(columns={df4.columns[0]: 'Close4'}, inplace=True)
#Long-term Holding
else:
start_date = form.cleaned_data['start_date']
end_date = form.cleaned_data['end_date']
df1 = web.DataReader(asset1, 'yahoo', start_date, end_date)
df2 = web.DataReader(asset2, 'yahoo', start_date, end_date)
df3 = web.DataReader(asset3, 'yahoo', start_date, end_date)
df4 = web.DataReader(asset4, 'yahoo', start_date, end_date)
df1 = pd.DataFrame({'Close1': df1['Close']})
df2 = pd.DataFrame({'Close2': df2['Close']})
df3 = pd.DataFrame({'Close3': df3['Close']})
df4 = pd.DataFrame({'Close4': df4['Close']})
portfolio_prices = df1.merge(df2, left_index=True, right_index=True) \
.merge(df3, left_index=True, right_index=True) \
.merge(df4, left_index=True, right_index=True)
print portfolio_prices
period = pd.bdate_range(start_date, end_date).date
weights = []
weights.append(weight1)
weights.append(weight2)
weights.append(weight3)
weights.append(weight4)
opti_model = MarkowitzOptimize(portfolio_prices, weights)
new_weights = opti_model.minimizeSharpeRatio()
attributes = list(portfolio_prices.columns.values)
return_prices = portfolio_prices / portfolio_prices.iloc[0]
return1 = return_prices[attributes].mul(weights).sum(1)
portfolio = pd.DataFrame(portfolio_prices)
attributes = list(portfolio.columns.values)
portfolio = portfolio[attributes].sum(1)
return2 = portfolio / portfolio.iloc[0]
ts_list = period.tolist()
date_string = [str(date) for date in ts_list]
args = {
'form': form,
'start_date': start_date,
'end_date': end_date,
'name1': name1,
'name2': name2,
'name3': name3,
'name4': name4,
'new_weights1': new_weights[0],
'new_weights2': new_weights[1],
'new_weights3': new_weights[2],
'new_weights4': new_weights[3],
'values1': return1.values.tolist(),
'values2': return2.values.tolist(),
'dates': date_string
}
return render(request, 'portimize/results.html', args)
def predict():
data = request.json
return predictions.predict(data)
description=
'Query for twitter mining, can be "search" or @username or #hashtag')
parser.add_argument('-q', '--query', help='Add your query', required=False)
parser.add_argument('-d', '--debug', help='Debug', required=False)
parser.add_argument('--all',
help='Model Selection',
required=False,
action='store_true')
args = parser.parse_args()
if __name__ == '__main__':
if args.query is None:
text = input("Enter Query: ")
else:
text = args.query
if args.debug:
debug = True
if args.all:
accuracy = False
else:
accuracy = True
tweets = findTweets(text)
predictions = predict(tweets[0], debug=debug, accuracy=accuracy)
for t, p in zip(tweets[1], predictions):
print('-' * 80)
print("Tweet:")
print('')
print(t, " \n### Predicted Sentiment[neg, pos]: ", p)
# print(predictions)
parsed = json.loads(lines[0])
return (parsed)
params = read_in()
#물가 예상
price_index = pr.index_predict(params["startdate"], params["year"],
params["price_index"])
price_index.name = "price_index"
solar_index = pr.index_predict(params["startdate"], params["year"],
params["solar_index"])
solar_index.name = "solar_index"
#수입예상 및 전체적인 지표구성
revenue = pr.predict(params["scene"], params["startdate"], params["year"])
revenue["rec_price"] = 100
revenue["days"] = revenue.index.day
start = pd.to_datetime(params["startdate"], format='%Y-%m-%d',
errors='ignore') #초기 기간 설정
startday = start.date().timetuple()
revenue.days[0] = revenue.days[0] - startday.tm_mday
revenue.days[revenue.shape[0] - 1] = startday.tm_mday
revenue = pd.concat([revenue, price_index, solar_index], axis=1)
revenue["generation"] = revenue['days'] * params['size'] * params[
'average_time'] * revenue["solar_index"]
revenue['smp_revenue'] = revenue['smp_price'] * revenue['generation']
revenue['rec_revenue'] = revenue['rec_price'] * revenue['generation'] * params[
'weight']
revenue["smp_revenue"] = revenue["smp_revenue"].astype(int)
# Init file
# All occurs here
# Execute this file tu start your machine
import sys
import training
import predictions
# Main Program
if __name__ == '__main__':
X_train, X_validation, Y_train, Y_validation = training.evaluate_data()
if 'plot' in str(sys.argv):
training.show_plots()
if 'predict' in str(sys.argv):
predictions.predict(X_train, X_validation, Y_train, Y_validation)
type="denac2",
_timebloc=timebloc,
n_scans_use=vd['num_scans_use'],
_shuffle=vd['shuffle_samples'],
_reduce=vd['reduced_timesteps'],
_avgpath=vd['path_to_avg'])
'''Callbacks and model fit'''
if (vd['UseObjective'] == 0 or vd['UseObjective'] == 1):
chkpt = ModelCheckpoint(vd['Modelpath'],
monitor='loss',
mode='min',
save_best_only=True,
verbose=1)
lrs = LearningRateScheduler(helpers.scheduler)
foo.model.fit(doo,
epochs=vd['epochs'],
verbose=1,
callbacks=[chkpt, lrs],
max_queue_size=vd['max_queue'])
'''Predictions'''
if (vd['UseObjective'] == 2):
predict(model=foo.model,
doo=doo,
type=vd['UseCase'],
predsavepath=vd['predsavepath'],
scaler=doo.scaler)
# '''Testing and Validation'''
# # test(doo, type = "denac" ,model_path = "post/model_den2.h5", item = 5772)