GE

Showcases a recommendation system using Amazon Personalize to generate real-time recommendations based on interactions by customers.

Project won the GE Hack'e'lth 2021 hackathon organised by GE Healthcare India and is waiting to be implemented at GE Healthcare.

Getting started

1. Deploy the backend on AWS

   1. Upload USERS.csv, ITEMS.csv and USER_INTERACTIONS.csv to an S3 bucket (Find these files in the shared GE box).

   2. Add a bucket policy allowing Amazon Personalize to access the contents.

      Bucket policy

      Replace BUCKET_NAME with your bucket name below.

      {
        "Version": "2012-10-17",
        "Id": "PersonalizeS3BucketAccessPolicy",
        "Statement": [
          {
            "Sid": "PersonalizeS3BucketAccessPolicy",
            "Effect": "Allow",
            "Principal": {
              "Service": "personalize.amazonaws.com"
            },
            "Action": ["s3:GetObject", "s3:ListBucket"],
            "Resource": [
              "arn:aws:s3:::BUCKET_NAME",
              "arn:aws:s3:::BUCKET_NAME/*"
            ]
          }
        ]
      }

   3. Create an IAM role with the AmazonS3ReadOnlyAccess permission. Steps 2 and 3 can also be automated by running initRole.ipynb

   4. In the Amazon personalize dashboard, create a dataset group and add the three datasets. The required shema for each of the datasets can be copied from below

      Users dataset schema

      {
        "type": "record",
        "name": "Users",
        "namespace": "com.amazonaws.personalize.schema",
        "fields": [
          {
            "name": "USER_ID",
            "type": "string"
          },
          {
            "name": "USER_HOSPITAL",
            "type": ["null", "string"]
          },
          {
            "name": "USER_ROLE",
            "type": "string",
            "categorical": true
          }
        ],
        "version": "1.0"
      }

      Items dataset schema

      {
        "type": "record",
        "name": "Items",
        "namespace": "com.amazonaws.personalize.schema",
        "fields": [
          {
            "name": "ITEM_NAME",
            "type": ["null", "string"]
          },
          {
            "name": "ITEM_FAMILY",
            "type": "string",
            "categorical": true
          },
          {
            "name": "ITEM_OVERVIEW",
            "type": ["null", "string"]
          },
          {
            "name": "ITEM_ID",
            "type": "string"
          }
        ],
        "version": "1.0"
      }

      Interactions dataset schema

      {
        "type": "record",
        "name": "Interactions",
        "namespace": "com.amazonaws.personalize.schema",
        "fields": [
          {
            "name": "USER_ID",
            "type": "string"
          },
          {
            "name": "ITEM_ID",
            "type": "string"
          },
          {
            "name": "ACTION",
            "type": "string",
            "category": true
          },
          {
            "name": "TIMESTAMP",
            "type": "long"
          }
        ],
        "version": "1.0"
      }

   5. Once the dataset import jobs have completed, create a solution with the arn:aws:personalize:::recipe/aws-user-personalization recipe.

   6. Create a campaign with the given solution version and copy the campaign ARN

   7. Create a lambda function with the code from getRecLambda.py. Copy the Campaign ARN to line 27.

   8. Attach an API gateway trigger to the function.

2. Start the web app

   Make sure you have NodeJS v16.6.0+ installed. Then, continue with the following steps:

   # Install dependencies
   npm install
   
   # Copy the example env file to .env.local, and fill in your endpoint URL from API gateway
   cp .env.local.example .env.local
   
   # The development server should be available on http://localhost:3000
   npm run dev

About

We submitted this project for GE Healthcare's GE Hack-E-LTH '21 hackathon and won 1st prize. Our team comprised of Joshua T, Samyuktha T H, Sandeep Rajakrishnan, and Vighnesh Shankar.