Create a Heroku App

[antoinecarme]

Need a RESTful API for PyAF and a Heroku demo.

https://pyaf.herokuapp.com/

[antoinecarme]

There is some preliminary working REST API app at https://pyaf.herokuapp.com/

It builds an ozone model by default.
try:
https://pyaf.herokuapp.com/models
and then :
https://pyaf.herokuapp.com/model/Ozone_Model_12/plot/all

[antoinecarme]

The REST API is implemented here : https://github.com/antoinecarme/pyaf/blob/heroku-tests/WS/WS_REST_Flask.py

[antoinecarme]

A new model can be added with a json body like this :

      {
        "CSVFile": "https://github.com/antoinecarme/TimeSeriesData/master/ozone-la.csv", 
        "DateFormat": "%Y-%m", 
        "Horizon": 12, 
        "Present": "1968-08", 
        "SignalVar": "Ozone", 
        "TimeVar": "Month"
        "Name": "new_name_for_the_model"

      }

POST at https://pyaf.herokuapp.com/model

[antoinecarme]

    "TrainOptionsHelp": [
      {
        "CSVFile": "A CSV file (URIs are also welcome!!!) containing a date column (optional, a integer sequence starting at zero is used if not present), and a signal column, for which the future values are to be predicted. ", 
        "DateFormat": "The format of the date column , if it is a physcial date/time/datetime column (iso : yyyy-mm-dd by default), empty otherwise", 
        "Horizon": "number of future time periods to be predicted. The length of a period is inferred from data (most frequent difference between two consecutive dates)", 
        "Name": "Name used to identify the model in the API", 
        "Present": "date/time of the last known signal value. Predictions start after this date/time", 
        "SignalVar": "Name of the signal column to be predicted", 
        "TimeVar": "Name of the date/time column"
      }

[antoinecarme]

The following plots re available :

        "AR": "https://pyaf.herokuapp.com/model/Ozone_Model_12/plot/AR", 
        "Cycle": "https://pyaf.herokuapp.com/model/Ozone_Model_12/plot/Cycle", 
        "Forecast": "https://pyaf.herokuapp.com/model/Ozone_Model_12/plot/Forecast", 
        "Prediction_Intervals": "https://pyaf.herokuapp.com/model/Ozone_Model_12/plot/Prediction_Intervals", 
        "Trend": "https://pyaf.herokuapp.com/model/Ozone_Model_12/plot/Trend", 
        "all": "https://pyaf.herokuapp.com/model/Ozone_Model_12/plot/all"

[antoinecarme]

The following SQL dialects are available :

        "Default": "https://pyaf.herokuapp.com/model/Ozone_Model_12/SQL/Default", 
        "mssql": "https://pyaf.herokuapp.com/model/Ozone_Model_12/SQL/mssql", 
        "mysql": "https://pyaf.herokuapp.com/model/Ozone_Model_12/SQL/mysql", 
        "oracle": "https://pyaf.herokuapp.com/model/Ozone_Model_12/SQL/oracle", 
        "postgresql": "https://pyaf.herokuapp.com/model/Ozone_Model_12/SQL/postgresql", 
        "sqlite": "https://pyaf.herokuapp.com/model/Ozone_Model_12/SQL/sqlite", 
        "sybase": "https://pyaf.herokuapp.com/model/Ozone_Model_12/SQL/sybase"

[antoinecarme]

Only the CSVFile is not optional!!!!

Set proper default values for all other settings.

Empty string and None are equivalent.
Default date format is ISO : 2016-12-31
When 'Present' is empty, use the whole dataset.
Default horizon is 1.

[antoinecarme]

A minimal model can be added with a json body like this :

  {
    "CSVFile": "http://chart.finance.yahoo.com/table.csv?s=GOOG&a=8&b=14&c=2015&d=9&e=14&f=2016&g=d&ignore=.csv"
  }

(Google stock quotes).

[antoinecarme]

sample usage (see WS/curl_tests.sh):

<script.sh>

PYAF_URL="http://pyaf.herokuapp.com/model"

CONTENT_TYPE="Content-Type: application/json"

DATA3="{"SignalVar":"Close", "Horizon":"21", "Present":"2016-01-01", "CSVFile": "http://chart.finance.yahoo.com/table.csv?s=GOOG&a=8&b=14&c=2015&d=9&e=14&f=2016&g=d&ignore=.csv\"}";

curl --header "$CONTENT_TYPE" -X POST --data "$DATA3" "$PYAF_URL"

</script.sh>

The previous curl command predicts three months (Horizon=21) of GOOG signal (yahoo finance API) starting from 2016-01-01.

[antoinecarme]

The API is working now.

Added a class:API label for future API-related issues.