Customer Segmentation

Term Project for University of Toronto's "SCS 3760 004 Cloud Technologies for Big Data, Machine Learning & Artificial Intelligence" course, Fall 2021.

Keywords: Google Cloud Platform; GCP; Vertex AI Workbench; Docker; Custom Container; Cloud Run; Artifact Registry; Cloud Build; Cloud Logging.

1. Setup Google Cloud

• Create a new GCP project;
• Enable these APIs: Cloud Run, Cloud Build, Cloud Logging, and Vertex AI;
• Create a bucket named gs://<PROJECT_ID>-bucket, where <PROJECT_ID> is to be replaced with th ID of the GCP project;
• Copy the unzipped content of the Customer Personality Analysis' dataset to gs://<PROJECT_ID>-bucket/custom-training/custsegm/data/marketing_campaign.csv.

2. Setup your working environment (Linux)

• Clone the repo:

  git clone https://github.com/msugar/customer-segmentation.git
  cd customer_segmentation
  

• Install and init the Google Cloud SDK. (Skip if you are in a Vertex AI Workbench's User-Managed Notebook.)
• Whenever you start a new working session, create/update your Python 3.7+ virtual environment:

  source setup.sh
  

3. Build and deploy the service

• Build the docker image locally:

  ./docker-build.sh cloudrun
  

• Submit the built image to Cloud Build -> Artifactory Registry -> Cloud Run:

  ./docker-deploy.sh cloudrun
  

• Test the service's health:

  curl <SERVICE_URL>
  

• Make a couple of predictions:

  curl -X POST -H "Content-Type: application/json" <SERVICE_URL> -d "@input.json"
  

4. Watch a video:

Video