pablomarin / Retail-Recommender

Recommender System in a docker image as API, for Retailers

The Accelerator powered by Azure OpenAI

The Retail Recommender Solution Offering unlocks data for clients to provide specific recommendations across item features such as; products, product categorites, sellers, customers, reviews on products, and more. This offering can be implemented as a POC in the clients environment with their own data in a 2 day workshop. The goal of the MVP workshop is to show/prove the value of a Recommender System built with the Azure Retail Recommender team, with your own data in your own environment. For more information on the 2 day workshop, click the powerpoint presentation below:

Link to Introductory Solution Workshop Deck

Click "view raw" to view powerpoint presentation

The Prerequisites

• Azure subscription
• Azure Machine Learning dedicated workspace
• Optional Databricks workspace

Prerequisites Client 2 Day Workshop

• Microsoft members need to be added as Guests in clients Azure AD
• A Resource Group (RG) needs to be set for this Workshop POC, in the customer Azure tenant
• The customer team and the Microsoft team must have Contributor permissions to this resource group
• A storage account must be set in place in the RG
• Datasets must be uploaded as CSV or Parquet files to the blob storage account, at least one week prior to the workshop date
• Azure Machine Learning Workspace must be deployed in the RG
• Optional but recommended – Databricks Workspace deployed in the RG

The Benefits

• In comparison with this MSFT accelerator, our Retail Recommender Solution Offering is much simpler to setup. It doesn't require Synapse or Spark while at the same time can handle more than 95% of the retail cases
• Uses not only the transaction log (Date, User_id, Item_id, Interaction), but can also use the Item master dataset and the user dataset in order to make quality predictions
• Leverages side features for training and for prediction. Can take in consideration side features for transactions , users and items.
• Can perform item similarity / basket analysis for upsell and cross sell
• Produces as output an API on a docker image. That can be place in any cloud provider or on-premises
• Contains deployment via Azure ML Services in:  Azure Container Instance (Test), and Azure Kubernetes Service (Production)}
• Also contains deployment for Azure Databricks MLFlow model serving
• API Features:  Top K recommendations, include or exclude set of items from the recommendation,  uses side features at prediction time, specify specific features at inference time,  allows querying similar items.

Getting Started and Process Overview

1. Create a compute instance within your Azure Machine Learning workspace (can use Standard DS11_v2)
2. Once your compute intance is created, launch your JupyterLab from your compute instance
3. Open the terminal within your JuypterLab and clone this repo:

# enter your own user directory below
cd /Users/pabmar/

# clone repo
git clone https://github.com/pablomarin/Retail-Recommender.git

# change directory to retail recommender
cd Retail-Recommender/

4. Once the repo is cloned, in your JupyterLab, navigate to the Retail Recommender Folder and open Notebook 0, the Data Download Notebook
5. Follow each notebooks 0-4 to complete the entire solution

Additional Steps for Data Download Notebook

1. Open Notebook 0 within your JupyterLab
2. Click on the Retail Rocket Dataset from the kaggle link (example dataset)
3. Download the dataset
4. Rename the file to "retailrocket-kaggle-data.zip"
5. Upload the file into your JupyterLab
6. Run the cell of code in Notebook 0
7. You should now have a new folder created called "data" with the files unzipped

Azure and Analytics Platform

The directions provided for this repository assume fundemental working knowledge of Azure Machine Learning Service and Azure Databricks MLFlow.

For additional training and support, please see:

1. Azure Machine Learning Services
2. Azure Databricks Services