`rel-hm` H&M Database

Database Description: The H&M relational database hosts extensive customer and product data for online shopping experiences across its extensive network of brands and stores. This database includes detailed customer purchase histories and a rich set of metadata, encompassing everything from basic demographic information to extensive details about each product available.

Database Statistics:

Num of Tables	3
Num of Rows	33,265,846
Num of Columns	37
Starting Time	2018-09-20
Validation timestamp	2020-09-07
Testing timestamp	2020-09-14
Time window	7 days

Database schema:

To load this relational database in RelBench, do:

from relbench.datasets import get_dataset
dataset = get_dataset("rel-hm")

References:

[1] H&M Personalized Fashion Recommendations Kaggle Challenge.

Dataset License: Non-Commercial Purposes & Academic Research.

Node Classification Tasks

`user-churn`

Task Description: Predict the churn for a customer (no transactions) in the next week.

Evaluation metric: AUROC

Node Regression Tasks

`item-sales`

Task Description: Predict the total sales for an article (the sum of prices of the associated transactions) in the next week

Evaluation metric: MAE

Link Prediction Tasks

`user-item-purchase`

Task Description: Predict the list of articles each customer will purchase in the next seven days.

Evaluation metric: MAP

rel-hm H&M Database