CausalStream : A General Architecture for Real‐Time Causal Inference on Data Streams
Concurrency and Computation: Practice and Experience · 2026-05 · 38 (10)
DOI: 10.1002/cpe.70756 ↗Kurz gefasst
Fachartikel in der Zeitschrift Concurrency and Computation: Practice and Experience, online erschienen am 18.05.2026. Verfasst von Omar El quammah und 3 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben.
Abstract (Originalsprache)
Real‐time causal inference on data streams is hindered by a fundamental bottleneck: traditional methods require O(N) reprocessing of historical data, making continuous treatment effect monitoring prohibitive at scale. We introduce CausalStream, a streaming architecture with an incremental doubly robust estimator achieving O(1) per‐event complexity through coordinated state management and sufficient statistics. A single‐node prototype validates the algorithmic core; the architecture is designed for distributed deployment with entity‐based partitioning and fault recovery. Evaluated on 350,000 retail transactions, CausalStream reduces 30‐day monitoring from 4500 to 143 s—a 31× improvement over batch reprocessing (3.8× on an equivalent windowed task)—while maintaining 2441 events/s throughput. Physical two‐node deployment confirms ATE consistency across nodes (deviation < 0.001), sustained throughput of 1933 events/s, and fault recovery with automatic reconnection. On benchmark datasets with known ground truth (IHDP, 1 M‐sample synthetic uplift), bias remains under 0.02 and throughput exceeds 7000 events/s. CausalStream bridges statistical causal inference and production streaming systems, enabling continuous causal intelligence for manufacturing, healthcare, and IoT.
Zitieren
APA
quammah, O. E., Weijun, C., Ouchkhi, O., & Darbinian, K. (2026). CausalStream : A General Architecture for Real‐Time Causal Inference on Data Streams. Concurrency and Computation: Practice and Experience, 38(10). https://doi.org/10.1002/cpe.70756
Vancouver
quammah OE, Weijun C, Ouchkhi O, Darbinian K. CausalStream : A General Architecture for Real‐Time Causal Inference on Data Streams. Concurrency and Computation: Practice and Experience. 2026;38(10). doi:10.1002/cpe.70756
BibTeX
@article{quammah2026,
author = {quammah, Omar El and Weijun, Cui and Ouchkhi, Ouahiba and Darbinian, Kristina},
title = {CausalStream : A General Architecture for Real‐Time Causal Inference on Data Streams},
journal = {Concurrency and Computation: Practice and Experience},
year = {2026},
volume = {38},
number = {10},
doi = {10.1002/cpe.70756}
}
EUVIMED