This document summarizes major improvements and fixes introduced in the Accelerate Product Similarity Optimization package release version.
|
Version |
1.1.0 |
|---|---|
|
Release Date |
Oct 21, 2024 |
New Features and Improvements
|
New Feature Description |
ID |
|---|---|
|
In the Definitions step on the Transaction Data Scope tab, settings of the "Use Transaction Source Data" checkbox are properly saved. |
PFPCS-7345 |
|
It is possible to compute product similarity for up to 1 million products. |
PFPCS-7390 |
|
It is now possible to evaluate the similarity on a subset of the initial data source to make "new products" be evaluated regarding their distance to the existing groups. |
PFPCS-7408 |
|
Group names are now displayed in singular only (to avoid duplication of words in group names). |
PFPCS-7490 |
|
In the Similarity Grouping Dashboard tab in the table “List of similarity groups”, the number of products has been added. |
PFPCS-7500 |
|
The default layout preferences are now part of the accelerator. |
PFPCS-7608 |
|
New products evaluation takes into account also numerical and categorical fields. |
PFPCS-7683 |
|
In the “New product scope” tab, there are two ways to assign new products from the previously computed most similar products: based on majority, or based on most similar products. |
PFPCS-7684 |
|
A Sankey chart displays how the product groups are extended with the new products affectations. |
PFPCS-7685 |
|
When installing the accelerator from PlatformManager, engine 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-7705 |
|
There is now a table “SimilarityTableExtended” as the “SimilarityTable”, including all products from the original scope and additional products. |
PFPCS-7998 |
|
In the Similarity Grouping Dashboard tab, "Product details" have been added. |
PFPCS-8141 |
|
Unified naming convention for the logics and elements has been applied. |
PFPCS-8149 |
|
productID and CoProductIDs were added as keys in the table similarityTableExtended. |
PFPCS-8279 |
Fixed Issues
|
Bug Description |
ID |
|---|---|
|
Names of tables and fields are not escaped in SQL queries. |
PFPCS-7580 |
|
In the Product Grouping step in the Product Overview tab, the table “Product Details” misses numerical attributes defined earlier. |
PFPCS-7763 |
|
The New Products step fails with an error "LoadData [via NewQueryUtils]" as there are missing the specific configurator and logics to deal with loading transactional data for new products. |
PFPCS-7930 |
|
The Product Grouping step fails with an "int64" value error as the column used as “ProductID” is automatically parsed into int64 if the ID is a numerical-like string. |
PFPCS-7962 |
|
When a product has only one similar product, there is an error. |
PFPCS-7987 |
|
When setting a filter for the New Product Scope tab in Step 4 Product Grouping, the New Product step fails. |
PFPCS-7988 |
|
Index error in Python calculation: the New Products step fails with an error "Length of values (237) does not match length of index (3)". |
PFPCS-8024 |
|
Group names include “Nan” for null values, which makes no sense. |
PFPCS-8104 |
|
There is an error “Singular matrix" when extending for more products. |
PFPCS-8132 |
|
Some numerical similarities are not computed normally. |
PFPCS-8163 |
|
Numerical similarity is computed incorrectly. |
PFPCS-8166 |
|
Number of points displayed in network chart "Similarity chart" has been limited to 40. |
PFPCS-8175 |
|
Numbers in the output tables are not properly formatted. |
PFPCS-8180 |
|
The accelerator computed pseudo-inverse of covariance matrix of numerical features for each set of candidates when labelling new products. Instead, it should be computed once, for numerical features of all existing products. |
PFPCS-8212 |
|
Unable to update the Model Class label to reflect the required prefix. |
PFPCS-8311 |