Easy Predictive Analytics can also use the REST API to update a prediction model (perform model retraining with the latest data). This process is executed by the series of API operations indicated below. 1.Retrain Data Upload API: Uploads data and executes retraining 2.Retrain Status Check API: Checks the status of retraining and retrieves a URL for downloading the results file after training 3.Retrain Model Switch API: Switches the model with API created to the retrained model

Retrain Data Upload API

Retrieves the URL to upload the data for retraining. When the file upload is complete, retraining is automatically executed.

URL

https://developer-api.easy-predictive-analytics.sony.com/v1/external/retrain_jobs/upload_url

The HTTP GET method is used.

Request

header

path parameter

Not used.

query parameter

body

Not used.

Response

body

The format is indicated below.

{
    "retrain_job_id": str,
	"upload_url": str
}

error message

Execution Example 1

An example of executing the Retrain Data Upload API in Python is indicated below.

import requests
import tempfile
import zipfile
import subprocess

api_url = 'https://developer-api.easy-predictive-analytics.sony.com/v1/external/retrain_jobs/upload_url'
api_key = 'YOUR_API_KEY'

headers = {'x-api-key': api_key}


# When Using a Training File Only
re_train_data_file_path = 'YOUR_RE_TRAIN_DATA_FILE_PATH'
re_train_data_file_name = 'YOUR_RE_TRAIN_DATA_FILE_NAME'

# request
params = {'filename': re_train_data_file_name}
response = requests.get(api_url, params=params, headers=headers)
response_json = response.json()
retrain_job_id = response_json['retrain_job_id']
upload_url = response_json['upload_url']

# upload_file
with open(re_train_data_file_path, 'rb') as f:
  data = f.read()
res = requests.put(upload_url, data=data)

# When Using Both a Training File and Evaluation File
re_train_zip_file_name = 'YOUR_RE_TRAIN_ZIP_FILE_NAME'
re_train_data_file_name = 'YOUR_RE_TRAIN_DATA_FILE_NAME'
re_train_data_file_path = 'YOUR_RE_TRAIN_DATA_FILE_PATH'
re_train_eval_data_file_name = 'YOUR_RE_TRAIN_EVAL_DATA_FILE_NAME'
re_train_eval_data_file_path = 'YOUR_RE_TRAIN_EVAL_DATA_FILE_PATH'

"""
dir_name
  |   |--train_src
  |   |   |--train_data.csv
  |   |--eval_src
  |   |   |--eval_data.csv
"""

with tempfile.TemporaryDirectory() as temp:
  zip_file_path = f'{temp}/{re_train_zip_file_name}'
	with zipfile.ZipFile(zip_file_path, 'w', compression=zipfile.ZIP_DEFLATED) as zf:
		zf.write(re_train_data_file_path, arcname=f'train_src/{re_train_data_file_name}')
		zf.write(re_train_eval_data_file_path, arcname=f'eval_src/{re_train_eval_data_file_name}')

  # request
  params = {'filename': re_train_zip_file_name}
  response = requests.get(api_url, params=params, headers=headers)
  response_json = response.json()
  retrain_job_id = response_json['retrain_job_id']
  upload_url = response_json['upload_url']

  # upload_file
  with open(zip_file_path, 'rb') as f:
    data = f.read()
  res = requests.put(upload_url, data=data)
  ret = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)

Execution Example 2

An example of executing the curl command from the command prompt of the Retrain Data Upload API is indicated below.

1.Launch the command prompt.

2. Change the view to UTF-8.
$ chcp 65001

3. Save a file indicating the file name in the UTF-8 format (with the name param.txt).

4. Make a request.
$ curl -o "response.json" -H "x-api-key:{YOUR_API_KEY}" -X GET -G https://developer-api.easy-predictive-analytics.sony.com/v1/external/retrain_jobs/upload_url --data-urlencode filename@"{Full path of file in 3}"
ex) $ curl -o "response.json" -H "x-api-key:xxxx-xxx-xxx-xxxx" -X GET -G https://developer-api.easy-predictive-analytics.sony.com/v1/external/retrain_jobs/upload_url --data-urlencode filename@param.txt

5. Check the response.json file and retrieve retrain_job_id and upload_url.

6. Upload the file.
$ curl -X PUT --upload-file {File path of data set for retraining} {upload_url retrieved in 5}