Skip to content
This repository was archived by the owner on Feb 13, 2026. It is now read-only.

README.md

Latest commit

 

History

History
76 lines (53 loc) · 2.99 KB

README.md

File metadata and controls

76 lines (53 loc) · 2.99 KB
Raw

Retail reorder prediction using Vertex AI

1. Overview

With the advent of cloud environments, the concept of huge capital investments in infrastructure in terms of capital and maintenance is a thing of the past. Even when it comes to provisioning infrastructure on cloud services, it can get tedious and cumbersome.

In this example, you will look at executing a simple PySpark code which runs on Serverless batch (a fully managed Dataproc cluster). It is similar to executing code on a Dataproc cluster without the need to initialize, deploy or manage the underlying infrastructure.

In this use case, we will be building a predictive model capable of using customer orders collected over time to predict which previously purchased products will be in a user’s next order.

Services Used

  • Google Cloud Storage
  • Google Cloud Dataproc
  • Google Cloud Bigquery
  • Google Cloud VertexAI

3. Permissions / IAM Roles required to run the lab

Following permissions / roles are required to execute the serverless batch

  • Viewer
  • Dataproc Editor
  • BigQuery Data Editor
  • Service Account User
  • Storage Admin
  • Notebooks Runner

4. Checklist

To perform the lab, below are the list of activities to perform.

1. GCP Prerequisites
2. Spark History Server Setup
3. Creating a GCS Bucket and Uploading Files
4. Creating a BigQuery Dataset

Note down the values for below variables to get started with the lab:

PROJECT_ID=                                         #Current GCP project where we are building our use case
REGION=                                             #GCP region where all our resources will be created
SUBNET=                                             #subnet which has private google access enabled
BQ_DATASET_NAME=                                    #BigQuery dataset where all the tables will be stored
BUCKET_CODE=                                        #GCP bucket where our code, data and model files will be stored
BUCKET_PHS=                                         #bucket where our application logs created in the history server will be stored
HISTORY_SERVER_NAME=                                #name of the history server which will store our application logs
UMSA_NAME=                                          #user managed service account required for the PySpark job executions
SERVICE_ACCOUNT=$UMSA_NAME@$PROJECT_ID.iam.gserviceaccount.com
NAME=<your_name_here>                               #Your Unique Identifier

5. Lab Modules

Following are the lab modules:

  1. Understanding Data
  2. Solution Architecture
  3. Execution
  4. Logging

The ways to perform the lab is,


6. CleanUp

Delete the resources after finishing the lab.
Refer - Cleanup