This short tutorial shows how to request data from the prediction service with python and requests (see also Access to REST-Services in Python).
For IDL you can take a similar approach by adapting these instructions to Access to REST-Services in IDL.
The Prediction Service
The prediction service represents a web interface which allows to request, insert and modify prediction data from the database. Operations are performed by sending URL requests, whereas each operation is well defined as a so called route. The property service comes along with a graphical user interface at http://localhost:8004/ui/ which provides visual access to all available routes. Hereby, all routes involving the request of data are enlisted under the View section.
For this tutorial we are using only two routes. One to verify wherever an algorithm configuration exists and one to request all available predictions within our database.
- /algoconfig/list
- /prediction/list
Implementation
Request
In our example we run a machine learning algorithm which produces a flare prediction to store within our database. Hereby, the algorithm consists of a training phase and a test or prediction phase. Within the training phase the algorithm learns and tunes its parameters which then can be stored within the database as a configuration for later use. Afterwards, within the test phase, we use this configuration to compute flare predictions which are also stored within the database.
For the following example we assume, as the two functions 'train_model' and 'test_model' are already defined within the python script. Otherwise, we refer to the following Howto: Ingest prediction data in database (REST API).
Given the two above two functions we could define our algorithm's workflow as follows:
import json
import requests
# setup environment
environment = {}
with open("params.json") as params_file:
environment = json.loads(params_file.read())
# download required regions and properties
response = requests.get(
'http://localhost:8002/region/%s/list?cadence=%s&time_start=between(%s,%s)'
% (
environment['algorithm']['dataset'],
environment['algorithm']['cadence'],
environment['algorithm']['time_start'],
environment['algorithm']['time_end']
)
).json()
if response['has_error'] == True or response['result-count'] == 0:
raise ValueError('No data found!')
else:
# FIXME: Bad Implementation
# if we already have an algorithm configuration stored within the database,
# we do not need to extract any training nor validation data.
(train_data, validation_data, test_data) = create_data_partitions(response['data'])
# setup model
model = DemoMLAlgorithm(environment['algorithm']['params'])
# check wherever we have to train our algorithm or if we can download an already existing configuration
response = requests.get(
'http://localhost:8004/algoconfig/list?algorithm_config_name=%s&algorithm_config_version=%s'
% (environment['algorithm']['cfg_name'], 'latest')
).json()
if response['has_error'] == False and response['result-count'] > 0:
# as we requested the latest configuration we expect only one result within 'data'
algo_cfg = response['data'][0]['config_data']
model.set_parameters(algo_cfg)
else:
train_model(
model, train_data, validation_data,
environment['algorithm']['max_epoches'], environment['algorithm']['batch_size'],
environment
)
# run algorithm
prediction = test_model(model, test_data, environment)
prediction_ids = [get_fc_id(row) for row in prediction]
# check wherever the new prediction was successfully stored within the database
response = requests.get(
'http://localhost:8004/prediction/list?prediction_fc_id=eq(%s)'
% (','.join(prediction_ids))
).json()
if response['has_error'] == True or response['result-count'] == 0:
raise AssertionError('No predictions found!')
else:
for prediction in response['data']:
print(prediction)
Source Code
Here you can download the full python source code.
For a more detailed implementation with dummy data you can download the python demo script.
...
This page was adopted from Request property data from database (REST API).