name string | domain string | description string | num_tables int64 | num_rows int64 | num_cols int64 | num_tasks int64 | tasks_binary_classification int64 | tasks_regression int64 | tasks_multiclass_classification int64 | tasks_multilabel_classification int64 | tasks_link_prediction int64 | start_timestamp string | val_timestamp string | test_timestamp string | size_gb float64 | license string | source_url string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
dbinfer-amazon | E-commerce (reviews) | Amazon from the 4DBInfer benchmark: a large product-review dataset linking users, products and reviews, used for rating prediction and user purchase/churn prediction. | 3 | 16,073,957 | 14 | 3 | 1 | 1 | 0 | 0 | 1 | 1996-06-25 00:00:00 | 2015-10-03 00:00:00 | 2015-12-30 00:00:00 | 5.7198 | see 4DBInfer / original sources | https://github.com/awslabs/multi-table-benchmark |
dbinfer-avs | Retail (Acquire Valued Shoppers) | Acquire Valued Shoppers (AVS) from the 4DBInfer benchmark: a retail dataset of customer transaction histories and promotional offers, used to predict shopper behavior such as offer repeat purchases. | 8 | 350,196,856 | 27 | 1 | 1 | 0 | 0 | 0 | 0 | 2012-03-02 00:00:00 | 2013-04-24 00:00:00 | 2013-04-30 00:00:00 | 3.2896 | see 4DBInfer / original sources | https://github.com/awslabs/multi-table-benchmark |
dbinfer-diginetica | E-commerce (sessions) | Diginetica from the 4DBInfer benchmark: an e-commerce dataset of user browsing and purchasing sessions over a product catalog (CIKM Cup 2016), used for click-through-rate and purchase prediction. | 12 | 97,258,572 | 32 | 2 | 1 | 0 | 0 | 0 | 1 | 2016-01-01 01:08:12.072000 | 2016-05-26 00:00:00 | 2016-05-31 00:00:00 | 0.2868 | see 4DBInfer / original sources | https://github.com/awslabs/multi-table-benchmark |
dbinfer-outbrain-small | Content recommendation | Outbrain (small) from the 4DBInfer benchmark: a content-recommendation dataset of document page views and promoted-content displays/clicks, used for click-through-rate prediction. | 9 | 4,213,483 | 34 | 1 | 1 | 0 | 0 | 0 | 0 | 1996-06-14 00:00:00 | 2016-06-24 00:00:00 | 2016-06-25 00:00:00 | 0.0501 | see 4DBInfer / original sources | https://github.com/awslabs/multi-table-benchmark |
dbinfer-retailrocket | E-commerce (behaviour) | RetailRocket from the 4DBInfer benchmark: an e-commerce dataset of visitor events (views, add-to-cart, transactions) over an item catalog, used to predict conversion (whether a viewed item is later purchased). | 7 | 24,885,583 | 18 | 1 | 1 | 0 | 0 | 0 | 0 | 2015-05-03 03:00:04.384000 | 2015-09-16 00:00:00 | 2015-09-17 00:00:00 | 0.4256 | see 4DBInfer / original sources | https://github.com/awslabs/multi-table-benchmark |
dbinfer-seznam | Digital advertising | Seznam from the 4DBInfer benchmark: a digital-advertising dataset from the Seznam.cz search engine, containing client prepaid-account charges and transactions, used to predict which service an account transacts on. | 4 | 2,688,678 | 14 | 2 | 0 | 0 | 2 | 0 | 0 | 2012-08-01 00:00:00 | 2015-04-01 00:00:00 | 2015-07-01 00:00:00 | 0.0211 | see 4DBInfer / original sources | https://github.com/awslabs/multi-table-benchmark |
dbinfer-stackexchange | Online community (Q&A) | StackExchange from the 4DBInfer benchmark: the Cross Validated (stats.stackexchange.com) community-Q&A dataset of users, posts, votes and badges, used to predict user churn and post upvotes. | 9 | 6,140,680 | 48 | 2 | 2 | 0 | 0 | 0 | 0 | 2009-02-02 00:00:00 | 2020-05-10 00:00:00 | 2021-05-18 00:00:00 | 1.0021 | see 4DBInfer / original sources | https://github.com/awslabs/multi-table-benchmark |
RelBench dbinfer datasets
This repository hosts the dbinfer family of relational datasets in the RelBench 3.0
manifest format, one subdirectory per dataset. The datasets originate from the
4DBInfer benchmark (data version
20240304), built directly from the original archives that dbinfer_bench itself
downloads (https://data.dgl.ai/mtbench/20240304-<name>.tar). Labels are the source's own, served as-is
(every task has kind: external).
Each subdirectory is a self-describing RelBench dataset (manifest.yaml + plain
db/*.parquet + tasks/<task>/); open its schema.svg for a zoomable
entity-relationship diagram.
Datasets
| dataset | domain | tables | tasks |
|---|---|---|---|
dbinfer-amazon |
E-commerce (reviews) | 3 | churn, purchase, rating |
dbinfer-avs |
Retail (Acquire Valued Shoppers) | 8 | repeater |
dbinfer-diginetica |
E-commerce (sessions) | 12 | ctr, purchase |
dbinfer-outbrain-small |
Content recommendation | 9 | ctr |
dbinfer-retailrocket |
E-commerce (behaviour) | 7 | cvr |
dbinfer-seznam |
Digital advertising | 4 | charge, prepay |
dbinfer-stackexchange |
Online community (Q&A) | 9 | churn, upvote |
Table counts include the single-column key tables materialized for the foreign-key targets
that 4DBInfer declares without a payload of their own (Item, Visitor, Customer,
Chain, Brand, Category, Company, Session, User, Orders, Token); see each
dataset card.
dbinfer-outbrain-smallhas almost no referential integrity, in the source. 4DBInfer subsampled its tables independently, so ~99.9% of its foreign keys -- and all but 58 of 69,543 distinct train entities of itsctrtask -- point at rows that were not kept. The previous revision hid this behind all-null keys. There is no full-sizeoutbrainarchive upstream. Its card has the numbers.
Several tasks derive their label from a column that is also in the database
(retailrocket/cvr <- View.added_to_cart, seznam/charge <- Dobito.sluzba,
seznam/prepay <- Probehnuto.sluzba, outbrain-small/ctr <- Click.clicked,
amazon/rating <- Review.rating). Each such task declares remove_columns; the column is
kept in db/ so the database stays faithful to 4DBInfer, and Dataset.get_db drops it, so
relbench.load_task(...) hands you a graph without it. If you read the parquet directly,
drop it yourself.
Loading
import relbench
ds = relbench.load_dataset("relbench/dbinfer/dbinfer-diginetica")
task = relbench.load_task("relbench/dbinfer/dbinfer-diginetica", "ctr")
db = ds.get_db()
train = task.get_table("train")
See the RelBench CONTRIBUTING guide for the manifest layout.
Provenance and revision history
Generated by
provenance/dbinfer.py,
which pins the sha256 of each source archive; verified by
provenance/check_dbinfer.py.
Port decisions -- which columns are kept, how implicit key domains are materialized, how
keys are reindexed -- are documented in that file's module docstring, and each dataset card
lists the payload columns it drops.
This collection was rebuilt from the original 4DBInfer archives. The previous revision
was derived from pre-built db.zip artifacts produced by the upstream
dbinfer-relbench-adapter export pipeline, which corrupted the data in ways that were not
visible from the schema:
- Every declared foreign key was 99.9-100% null. The adapter validated foreign keys
against
len(parent_table)where the parent was a{column: array}dict -- i.e. against the parent's column count -- so all larger key values were set toNaN. None of the seven databases could be joined. - Real primary keys were overwritten with
np.arange(n)before that validation, so the key correspondence was already gone. - Foreign keys to non-materialized key domains were dropped, amputating the schema
(
retailrocket/Viewwas left with no foreign keys at all). - Task entity ids were remapped through a task-local mapping unrelated to the database's, so label rows did not reference the database.
- Task time columns were destroyed (
1970-01-01 00:00:00.000000022) by casting the source's integer-valued columns withastype('datetime64[ns]'). - Classification targets were silently relabelled by sorted-string order, and
task_typewas inferred from target cardinality -- turningamazon/rating, a 4DBInfer regression/RMSE task, into multiclass, and both retrieval tasks into multiclass. - Undeclared payload columns were published, including
Posts.Score(from whichstackexchange/upvote's label is derived),History.repeater(avs/repeater's label itself), and full-history aggregates likeUsers.Reputation/UpVotes/Views. val_timestamp/test_timestampfell after the end of the data, soget_db(upto_test_timestamp=True)trimmed nothing and gave no temporal protection at all (dbinfer-retailrocketclaimed2015-09-21against a last event of2015-09-18;dbinfer-digineticaclaimed2016-11-12;dbinfer-seznam2015-10-04against2015-10-01). They also did not bracket the source's own splits -- seznam's val labels start at2015-04-01and its test labels at2015-07-01.
The current revision fixes all of the above: foreign keys resolve, primary keys are dense,
implicit key domains are materialized as tables, task entity columns index their entity
table, time columns are real timestamps, targets keep their source values, task_type
follows the source, and each task declares remove_columns for any database column its
label is derived from. Results computed against the previous revision are not
comparable.
Citation
These datasets are from the 4DBInfer benchmark. If you use them, please cite:
@inproceedings{wang2024fourdbinfer,
title = {{4DBInfer}: A {4D} Benchmarking Toolbox for Graph-Centric Predictive Modeling on Relational Databases},
author = {Wang, Minjie and Gan, Quan and Wipf, David and Cai, Zhenkun and Li, Ning and Tang, Jianheng and Zhang, Yanlin and Zhang, Zizhao and Mao, Zunyao and Song, Yakun and Wang, Yanbo and Li, Jiahang and Zhang, Han and Yang, Guang and Qin, Xiao and Lei, Chuan and Zhang, Muhan and Zhang, Weinan and Faloutsos, Christos and Zhang, Zheng},
booktitle = {Advances in Neural Information Processing Systems 37 (NeurIPS 2024) Datasets and Benchmarks Track},
year = {2024}
}
- Downloads last month
- 919