Menu Close

Main considerations:

[ultimate_spacer height=”10″ height_on_tabs=”10″ height_on_tabs_portrait=”10″ height_on_mob_landscape=”10″ height_on_mob=”10″]
  • Ideally the user should have basic skills on Amazon Elastic Container Service (ECS).
  • The files “metadata” and “data” represent a scenario of 50 products sales, with a daily frequency and integer data type.
  • The output for this example is a csv file with the forecast for 5 days for each one of the series.
  • The data provided by the user must keep the format of the data and metadata files.
  • The elapsed time for the forecast delivery are variable, but it is estimated up to 40 minutes with the recommended settings in the following instructions.

How can you use it?

Next you will find the steps to follow in AWS for the correct use of the product:

1. Get the access_key and secret_key associated to the AWS account. If you already have them, follow step 2. Otherwise, go to “My Security Credentials”

and create a new access_key, where a new secret_key will be generated

2. In S3 create a bucket with a specific name (it is recommended to use the company name – DNI) and upload to it the data and metadata files.

3. Go to Amazon ECS and click on “Gest Started”

Click on “Configure” to define a container

Then name the container, enter the address of the image-container and add the parameter memory limits to 512. You must delete the port mappings (click on the blue cross).

Then click on “Advanced container configuration”, in Environment Variables, add these 3 variables:

  • access
  • secret
  • bucket

Then check “value” to enter in “key” the following information:

  • access = enter access_key
  • secret = secret_key
  • bucket = the name of the bucket created in step 2

Then click on “Update”

4. Click on “Edit” to edit the task parameters.

5. In the memory configuration add the following parameters:

Then “Save” and “Next” until you get to the “cluster name” option, where a random name will be assigned. Finally “Next” and “Create”.

6. With this the product will be executed, the forecast outputs will get to the bucket created, generating a folder named “output_forecast” with the forecast files.

7. To return to the product you only need to update the task, by clicking on “Update”, then on “Skip review” and finally in “Update service” (no other parameter must be changed). Before this you must change the data in the bucket, so the forecast engine recognizes the new inputs.