This document summarizes major improvements and fixes introduced in the Accelerate Product Recommendation Optimization package release version.
|
Version |
1.0.2 |
|---|---|
|
Release Date |
Jul 18, 2023 |
New Features and Improvements
|
Description |
ID |
|---|---|
|
Python engine is now automatically deployed to the partition (if not already present) together with the accelerator. |
PFPCS-6938 |
Fixed Issues
|
Bug Description |
ID |
|---|---|
|
Product Recommendation step fails when Pricing Date is used for Basket ID but this field is not unique for each product-customer pair. To fix this issue, this corner case was implemented: transaction date is used for Basket ID and a customer orders the same product multiple times on a single date. |
PFPCS-6808 |
|
The Product Recommendations step fails with an error "Input X contains NaN. AgglomerativeClustering does not accept missing values encoded as NaN natively..." This happens when the selected data only contains a single customer and model generated segments are selected. To fix this issue, recommendations for a single customer are now produced. |
PFPCS-6811 |
|
When using model generated customer segments, you can get an error "cannot extract more clusters than samples". |
PFPCS-6824 |
|
Product Recommendation's SegmentMapTable DMT fields are not classified (dimension vs. measure) and thus cannot be used by the Data Table input. |
PFPCS-6831 |
|
The Product Recommendation step fails with an error "duplicate key value". |
PFPCS-6862 |