Optimization - Product Recommendation 2.0.0

This document summarizes major improvements and fixes introduced in the Accelerate Product Recommendation Optimization package release version.

Version

2.0.0

Release Date

Dec 28, 2023

New Features and Improvements

Improvement Description

ID

Improvement Description

ID

Model setup has been split into 3 steps: Definition (with Transaction Source and basic mapping), Configuration (with thresholds for recommendations) and Product Recommendations (remains the same).

PFPCS-6996

To define easily product recommendations, you can now turn on/off each type by using a checkbox in the Configuration step.

PFPCS-6997

New recommendation types have been added to product recommendations”

  • Similar products

  • Up-sell

  • Down-sell

  • Similar from another brand

PFPCS-7035
PFPCS-7043
PFPCS-7047
PFPCS-7051

An "overall score" has been added in order to get a score that combines all types of recommendations. This score uses weights that are defined by the recommendation type. 

PFPCS-7053

Recommendation types existing in version 1.0 match requirements introduced in this version. They are computed only if selected, otherwise the calculation is skipped.

PFPCS-7226

Model evaluation was updated: the ProductRecommendation element was modified to have a method for each of the newly added recommendation types. The AggregatedRecommendation element was added for evaluation method aggregating all types of recommendations into a single output. The EvaluationUtils element was added for functions used in the two above mentioned elements.

PFPCS-7409

Each recommendation type is now presented by a single node in the graph on the Summary tab. All new recommendation types are added, with the number of recommendations and number of co-products for which they are enabled.

PFPCS-7563

The Summary tab of the results step shows all recommendation types, with a number of unique product IDs and total count of recommendations.

PFPCS-7591

Default preferences for layout are now part of the accelerator release.

PFPCS-7609

When installing the accelerator from PlatformManager, engines configuration steps are now the first steps of the accelerator installation. This is to prevent errors when trying to run a model while the accelerator installation has not been completed.

PFPCS-7704

All scores returned by the evaluation are now weighted and they are added to each recommendation type table and presented at the overview tab.

PFPCS-7791

Product Recommendation models run now on partitions using Greenplum database.

PFPCS-7801

With version 2.0 the Model Class, tabs, steps and logics have been renamed to clearly indicate that they belong to the new version.

PFPCS-7806

Fixed Issues

Bug Description

ID

Bug Description

ID

When you use productId in the Product Name field, you get the following error: "column reference "productid" is ambiguous".

PFPCS-6564

When the source Datamart contains records sharing the same product/customer IDs but with different product/customer names, LoadRecommendations calculation fails with an error.

PFPCS-7266