Tagged articles
traffic allocation
16 articles · Page 1 of 1
The article presents a CIKM‑2024 paper that introduces MODRL‑TA, a multi‑objective deep reinforcement learning system combining multi‑objective Q‑learning, a cross‑entropy‑based decision‑fusion algorithm, and a progressive data‑augmentation pipeline to dynamically allocate search traffic on JD.com, with both offline and online experiments showing substantial gains in CTR, CVR, and overall platform performance.
Multi-Objective OptimizationOnline A/B Testingcross-entropy method
0 likes · 14 min read

MODRL‑TA is a multi‑objective deep reinforcement learning framework that unites independent Q‑learning agents, a cross‑entropy‑based decision‑fusion module, and progressive data‑augmentation to overcome cold‑start and multi‑objective trade‑offs in e‑commerce traffic allocation, delivering up to 18% more impressions, 4% higher CTR and 5% higher CVR in live tests.
deep learninge-commercemulti-objective
0 likes · 14 min read

This article explains how Huolala's data‑science team tackles interference between multiple time‑slice experiments by using city‑level isolation, nested experiment planning, and bias‑variance trade‑offs, providing detailed guidelines, recovery cycles, and case studies to maximize traffic utilization and experimental reliability.
A/B testingbias‑varianceexperiment design
0 likes · 11 min read

This article explains the challenges of A/B testing in two‑sided advertising markets and presents several experimental designs—including four‑cell traffic experiments, counterfactual interleaving, joint sampling, and simulation systems—illustrated with Tencent’s practical implementations to mitigate interference, spillover, and competition effects.
Advertisingad experimentscounterfactual interleaving
0 likes · 15 min read

Xiaohongshu’s new “traffic‑optimal + dynamic decision” framework models splash‑screen ad allocation as a linear‑programming problem with volume guarantees, continuously adjusts weights via feedback, and pre‑computes cached decisions to preserve fast app startup, thereby boosting click‑through rates while meeting delivery commitments.
CTR improvementDynamic DecisionLinear Programming
0 likes · 14 min read

The article presents the design, scientific methodology, system architecture, and case studies of Yanxuan's A/B testing platform, detailing how statistical principles, automated tracking, traffic allocation models, and unified reporting accelerate decision‑making and reduce development effort in e‑commerce experiments.
A/B testingautomationdata pipeline
0 likes · 15 min read

This talk by JD Retail's Zhao Yu explains how reinforcement learning is modeled and deployed for large‑scale traffic control during major sales events, detailing system architecture, reward design, offline simulation, model upgrades, and future research directions.
JD.comOnline AdvertisingRL modeling
0 likes · 20 min read

This article details the design and implementation of a transparent, explainable traffic‑decision system for an e‑commerce platform, covering background challenges, directional principles, selection and targeting methods, PV value estimation, allocation algorithms, and the supporting data‑engineering and visualization infrastructure.
PV estimationProduct Selectionalgorithmic fairness
0 likes · 22 min read

Huolala’s A/B testing platform, serving over 200 business scenarios and thousands of experiments, combines a three‑stage workflow with a modular architecture—including a configuration console, SDK for traffic routing and data collection, and a robust data‑recovery pipeline—while offering five distinct allocation algorithms to ensure scientific experiment results.
A/B testingalgorithm designexperiment platform
0 likes · 17 min read

This article explains how the Picasso platform provides a standardized, high‑efficiency A/B testing pipeline for front‑end developers, covering traffic allocation algorithms, flow reuse, orthogonal and mutually exclusive experiments, multi‑scenario rules, custom metrics, reporting, and future enhancements.
A/B testingcustom metricsexperiment platform
0 likes · 12 min read

The article compares non‑intervention (natural) traffic, where models autonomously maximize UV, with intervention (allocation) traffic that fine‑tunes re‑ranking to meet short‑term business goals, outlines objectives of balancing immediate profit and long‑term value, and presents two technical solutions—an ML‑plus‑OR integer‑programming model and a PID‑based control loop—for real‑time traffic allocation.
Operations ResearchPID ControlReal-time Decision
0 likes · 9 min read

The paper presents an AB‑testing platform that structures experiments into scenes, buckets, layers and traffic, uses a salted‑hash based allocation and a two‑step “multi‑withdraw‑fill” algorithm to adjust percentages while preserving user‑experiment stability, and describes a lightweight, cache‑centric system architecture with staggered config reloads and safeguards against database spikes and zombie nodes.
AB TestingAlgorithmexperiment platform
0 likes · 11 min read

This article explains how 58.com designs a three‑layer recommendation system for its homepage, tackles challenges of multi‑business fusion, interest modeling, traffic allocation, and dynamic refresh, and presents a step‑by‑step optimization pipeline that improves CTR and diversity.
AIdynamic-refreshinterest modeling
0 likes · 17 min read

The article details the design principles, layered traffic allocation model, architecture, data processing pipeline, and operational experience of the AB experiment platform used at Beike Zhaofang, highlighting its web, API, and storage layers, gray‑release capabilities, current limitations, and future improvements.
AB Testingbackend architectureexperiment platform
0 likes · 15 min read

The article describes the motivation, layered design, traffic allocation, statistical validation methods, and system architecture of a scalable A/B testing platform that enables multiple concurrent experiments while ensuring independent traffic segmentation and reliable data analysis for product growth.
A/B testingconfidence intervalexperiment platform
0 likes · 12 min read

This article explains the online side of recommendation systems, covering data flow, traffic allocation, AB testing, the gateway, recall mechanisms, and real-time ranking, while also describing how offline-trained models are deployed as binary weight files for live serving.
AIRecommendation Systemonline ranking
0 likes · 7 min read
