Recommendation model

Calculates best suitable products of a eCommerce client for a Customer based on Transaction history

Datasets used to feed the model

• Products (.txt)
• Product Categories (.tsv)
• Transactions (complex JSON)

Technologies used

• Databricks (Spark 2.4.6)
• Python 3.6
• Google Cloud Platform (GCP)
• Big Query
• Rest API calls

Model breakdown:

• The model feeds on products, categories and transaction datasets and identifies highly purchased products(and corresponding categories) based on transaction history.
• The transaction dataset doesn't have driving column; made use of Python uuid to generate a random unique column values to perform self join on transaction dataset to identify highly transacted products and categories.
• Retrieve test service engine credentials by making REST calls from GCP using service account details provided from input file.
• Prepared JSON request out of identified highly transacted dataset to perform REST call to TEST service engine hosted on GCP to validate the model performance.

Future Scope:

• Self Learning - scikit library implementation.
• Integrate Email notification about recommended products to customers.

Branching

Latest emhancements will be updated to Master branch for release.

1. master branch

   1. merge needs Pull Request review/approval.
   2. Once reviewed and merged with develop, raise Pull Request for master for enhancement to be made.

2. develop branch

   1. Create Enhancement wise branches out of it.
   2. Work enhancements wise contribution.
   3. Push latest code with Pull Requests and get reviewed.
   4. merge needs Pull Requests review/approval.