By moving real‑time recommendation computation to users’ devices and using a two‑stage edge‑cloud CTR model to gate Xianyu’s Mario feature, the system cut server requests by 28%, boosted Mario’s click‑through rate by 31% and lifted overall conversion by 10% while preserving user privacy.
The article examines Didi’s “Guess Where You’re Going” feature, describing its product benefits, the contextual data used, and a simple Bayesian Gaussian model that predicts a user’s destination based on time, departure location, and weekday, while also discussing its limitations and potential improvements.
This article excerpt explains the structure and training of feed‑forward neural networks, illustrates their use in neural language models, describes deep structured semantic models for ranking tasks, and details two‑stage recommendation systems such as YouTube, covering both theoretical formulas and practical deployment considerations.
This article reviews the TransFM model, which combines the translation‑based sequential recommendation approach (TransRec) with factorization machines (FM), explains its formulation, optimization via sequential Bayesian personalized ranking, and demonstrates its superior performance on Amazon and Google Local datasets compared with several baselines.
This article presents a comprehensive overview of Club Factory's recommendation system, covering its product recommendations, key challenges such as user and item cold‑start, the modular architecture, detailed recall and ranking processes, and practical considerations for deployment in e‑commerce.
This article shares JD's e‑commerce recommendation system practice, covering the overall online/offline architecture, recall and ranking modules, real‑time feature and model updates, multi‑objective and diversity strategies, first‑stage index‑based ranking, KNN recall, and a layered experiment platform for rapid iteration.
This article summarizes Alibaba’s twelve KDD 2019 papers, covering advances in long‑sequence CTR modeling, fashion recommendation, sponsored search, exact‑K recommendation, meta‑learning, transfer learning, scalable graph convolution, heterogeneous graph neural networks, knowledge‑driven product description, and related workshops, highlighting both algorithmic innovations and industrial deployments.
The article details JD.com’s journey from rule‑based product recommendations in 2012 to a sophisticated, AI‑driven personalized recommendation system, describing its multi‑screen product types, data collection, offline and online computation pipelines, and the modular architecture of its recommendation engine.
This article introduces recommendation ranking, breaks down its components, explains how to build ranking models, and provides additional insights, while also presenting the author's background and related job opportunities in the e‑commerce recommendation field.
The article traces the historical development of recommendation systems from early manual and hot‑ranking methods through natural ranking and machine‑learning‑based scoring, discusses the Matthew effect and its mitigation via randomization, multi‑objective weighting, and pipeline architectures, and outlines modern personalization and recall strategies for e‑commerce platforms.
This article presents the design, optimization, and deployment of a deep‑learning based ranking pipeline and model for NetEase News, covering offline and online pipelines, feature‑processing enhancements, custom TensorFlow operators, modular model architectures, and performance improvements demonstrated in production.
This article explains how Beike, China’s largest residential service platform, leverages its massive house, client, and text data to design a multi‑layered tag architecture, detailing data sources, tag construction methods—including classification, keyword, geographic, anonymous topic, and temporal tags—and their application to improve personalized house search and recommendation.
This article describes how Tujia applied embedding techniques, particularly a Skip‑Gram model, to build an item‑to‑item similarity recommender for low‑frequency, highly personalized homestay listings, detailing the data preparation, model architecture, training process, evaluation results, practical improvements, and future directions.
This article introduces the "Kelu Fang" product, which leverages knowledge graph, NLU, and ranking technologies to enhance real‑estate search by adding commute‑based filtering and a local view of surrounding facilities, and discusses its architecture, implementation details, and future improvement directions.
This article examines the role of search as a user guidance channel in e-commerce, outlining product requirements, user flow stages, and various algorithmic solutions—including query understanding, suggestion, rewriting, retrieval, and ranking—while also comparing implementations across major Chinese platforms.
This article explains how Alibaba’s Search & Recommendation team constructs a large‑scale e‑commerce concept graph—defining e‑commerce concepts, mining them from queries and titles, building an ontology, linking concepts to entities, and applying the graph to improve personalized search and recommendation.
This article explains the fundamentals of the Word2Vec SGNS model, details its hyper‑parameters and training tricks, and demonstrates how customized embeddings are built for rental‑listing and news‑article recommendation, covering data preparation, objective‑function redesign, evaluation, and deployment in both recall and ranking stages.
This article presents the design and technical details of 360 Easy Placement, an AI‑driven end‑to‑end advertising platform that automates creative generation, fast review, and optimization, addressing the challenges faced by small‑and‑medium advertisers through data‑rich models, multi‑task learning, and intelligent scene recommendation.
This article presents a comprehensive overview of Zhihu's recommendation page ranking system, covering its request flow, historical ranking evolution, feature design, deep learning models, multi‑task CTR optimization, practical engineering insights, current challenges, and future research directions such as reinforcement learning.
This article explains what a tourism knowledge graph is, discusses its architecture, construction methods, practical applications such as QA and recommendation, and explores future directions integrating knowledge graphs with deep learning and multi‑domain fusion.
The 2018 Suning Search & Recommendation Technology Conference in Nanjing gathered over 400 industry experts to discuss search engine evolution, recommendation algorithm models, multi‑source data fusion, multimedia video retrieval, and AI‑driven advertising, highlighting practical implementations and future research directions.
This article outlines the background of Ele.me's recommendation business, details the evolution of its recommendation algorithms from rule‑based models to deep learning and online learning, and explains the practical implementation of real‑time data pipelines, feature engineering, model training, and deployment.
This article introduces the ListNet listwise ranking algorithm, explains its theoretical foundations and loss function, presents a Python implementation with gradient computation, and demonstrates its superior performance over pointwise and pairwise models on public benchmarks and an internal automotive dataset for predicting users' intended car series.
This article reviews the evolution of Alibaba’s e‑commerce search and recommendation technologies, detailing Taobao’s unique challenges, the shift from rule‑based retrieval to large‑scale machine learning, real‑time online learning, deep learning and intelligent decision‑making, and outlines future directions toward cognitive intelligence.
This article examines the distinctive features of tourism-domain knowledge graphs, outlines methods for constructing them from internal and external data sources, and explores their practical applications such as question‑answering bots, personalized recommendation, and advanced search within the online travel sector.
The YMM-TECH algorithm competition final, held at Nanjing University of Posts and Telecommunications, presented a logistics recommendation problem that leverages driver behavior data, evaluates solutions using ranking‑accuracy metrics with position‑weighted scores, and provides detailed formulas, examples, and data‑driven recommendations for 20 cargo items per driver.
This article details the evolution of Alibaba's short‑video feed ranking system, from a Wide&Deep CTR model to a hybrid Factorization‑Machine and Gradient‑Boosted‑Tree approach, describing feature engineering, model architecture, experimental results, lessons learned, and future directions toward duration‑based relevance.
This article introduces the Wide&Deep, PNN, DeepFM, and a novel Wide‑ResNet model applied to Alibaba’s 1688 “You May Like” recommendation, describes the system architecture, training data, experimental results showing AUC improvements with batch normalization, and shares practical tuning insights.
Xianyu’s personalized recommendation system struggles with tiny, fast‑turnover item pools because traditional X2I matrices provide insufficient recall, so the team introduced pool‑specific pre‑filtering, high‑dimensional vector search, and a real‑time search‑engine recall, the latter boosting clicks by 14 % and transactions by 0.14 %.
This article introduces Wide&Deep, PNN, DeepFM, and a novel Wide‑ResNet model for Alibaba’s 1688 “Guess You Like” recommendation, explains the underlying feature services and real‑time scoring pipeline, presents offline experiments showing AUC gains with batch normalization, and shares practical tuning insights.
This article explains how short‑video platforms enhance recommendation quality by combining click‑through‑rate models with multi‑objective optimization of watch time and completion rate, using sample reweighting and weighted logistic regression to balance perceived and real relevance while improving offline AUC and online user engagement.
This article explains how JD.com evolved its recommendation system from simple product suggestions to a sophisticated, multi‑screen, multi‑type personalized engine using big‑data collection, real‑time behavior tracking, machine‑learning models, and a modular architecture that boosts conversion and user experience.
This article summarizes Sogou's information‑feed recommendation system, covering the architecture from data collection and NLP processing to recall, ranking, and feedback, and detailing the classification, tagging, keyword extraction, and various recall and ranking models such as FastText, TextCNN, collaborative filtering, and wide‑and‑deep learning.
The article outlines the development, multi‑screen deployment, system architecture, data platform, and core recommendation engine of JD.com’s e‑commerce recommendation platform, highlighting how big‑data and AI techniques enable personalized product, activity, and content suggestions across various user touchpoints.
This paper proposes a novel attention‑based framework that groups and encodes heterogeneous user behavior sequences into separate semantic subspaces, applies self‑attention to capture inter‑behavior influences, and demonstrates faster training and comparable or improved recommendation performance across multiple tasks and datasets.
This article reviews the evolution of recommendation system architecture for C2C e‑commerce platforms, tracing stages from simple offline‑online pipelines through granular feed‑flow improvements, real‑time processing, and machine‑learning‑driven models, while highlighting user‑profile construction, challenges, and best‑practice guidelines.
This article introduces the most commonly used algorithms in engineering—recommendation, optimization, estimation, and classification—explains their typical application scenarios, and discusses various testing methods and evaluation metrics such as offline experiments, user surveys, A/B testing, and performance indicators like accuracy, coverage, diversity, and robustness.
Alibaba’s recent Search and Computing Open Day revealed how AI, big‑data algorithms and multi‑objective optimization are reshaping its e‑commerce ecosystem, boosting commercial efficiency, supporting diverse merchants, and driving innovative retail experiences for billions of users.
This article traces the evolution of Weibo's recommendation system architecture across three major phases—independent 1.0, layered 2.0, and platform‑centric 3.0—detailing the environmental drivers, technical components, advantages, drawbacks, and key outcomes of each stage.
This article summarizes a AAAI 2018 paper that introduces a privacy‑preserving, decentralized matrix‑factorization approach for point‑of‑interest recommendation, detailing its problem definition, model design, random‑walk based user interaction, experimental evaluation on Foursquare and Alipay datasets, and future research directions.
This article details Ctrip's large‑scale personalized recommendation system, covering the data sources, recall strategies, ranking models, feature engineering techniques, and future directions for improving recommendation quality in the travel domain.
This article explains the objectives of recommendation systems, compares popular recommendation approaches, details the components and algorithms of personalized recommendation pipelines, and discusses practical challenges such as real‑time processing, freshness, cold‑start, diversity, content quality, and surprise handling.
This article presents a comprehensive NLP‑driven text‑mining workflow for tourism comment data, covering data cleaning, word2vec training, keyword extraction, sentiment analysis, ranking, and a lightweight architecture that enables fast, accurate recommendation of scenic spots for users.
This article describes how Meituan‑Dianping leverages deep learning, especially the Wide & Deep model, to improve its recommendation system by addressing business diversity, user context, feature engineering challenges, optimizer and loss function choices, and presents offline and online experimental results showing significant CTR gains.
58.com’s intelligent recommendation system, evolving from a C++ monolith in 2014 to a Java-based micro‑service platform, integrates multi‑layer data processing, diverse recall and ranking algorithms, and a robust microservice architecture to deliver personalized listings across housing, jobs, cars, and more.
This article presents a comprehensive overview of personalized recommendation systems, covering their application scenarios and value, user profiling, core algorithms such as content‑based and collaborative filtering, system architecture, performance and effect optimization techniques, and practical Q&A insights.
The article details Ctrip CTO Gan Quan’s insights on how the travel platform leverages a comprehensive big‑data infrastructure, AI‑driven algorithms, and real‑time user behavior tracking to deliver personalized travel recommendations, improve conversion rates, and shorten user decision cycles across multiple business lines.
The article details how Youku Tudou redesigned its video recommendation architecture, moving ranking from offline to online processing, outlining the comparative architecture, benefits, challenges, feature handling, offline evaluation methods, and weight‑fusion techniques that enabled a successful launch after two months of development.
This article details how Ele.me built a rapidly iterating recommendation system, covering model ranking architectures (single, linear, multi), online feature computation pipelines, feature management, and shuffling logic to balance algorithmic relevance with user perception, providing practical insights for large‑scale personalized services.
At Alibaba's 2016 Double 11 Tech Forum, researcher Xu Yinghui presented how online AI technologies enhance search and recommendation on the e‑commerce platform, turning massive user behavior data into actionable insights that improve traffic allocation and maximize welfare for consumers, sellers, and the platform.
This article describes Qunar's personalized demand prediction system for the "Guess You Like" card, detailing how user‑demand associations are mined via rule engines, collaborative filtering, LBS and manual rules, and how ranking models evolve from subjective Bayes to RankBoost and LambdaMart, with experimental evaluation and future LSTM plans.
The article describes how YHD's precision recommendation team built a real‑time, highly available, traceable general recommendation platform, detailing its background, overall architecture, visual configuration and traceability subsystems, and reporting significant improvements in development speed, reuse and user satisfaction.
This article describes how Qunar replaced a low‑click hot‑words card with a personalized “You May Like” recommendation card, detailing data collection, rule‑based and collaborative‑filtering association methods, learning‑to‑rank models (subjective Bayes, RankBoost, LambdaMart), training‑sample strategies, online experiments, evaluation metrics, and future plans including LSTM‑based sequence modeling.
This article details 58 Daojia's data‑driven approach to building user‑portrait tags, covering tag construction, evaluation, and practical applications such as personalized recommendations, anti‑fraud measures, coupon distribution, and dynamic pricing, while outlining the underlying big‑data architecture and technical challenges.
This article outlines the design of both system-wide and personalized recommendation modules for e‑commerce platforms, explains three recommendation approaches (demographic, content‑based, collaborative filtering), details the implementation of Mahout’s collaborative‑filtering algorithm with Java code, discusses data sources, technology stack, algorithm choices, and solutions to the cold‑start problem.
This article explains why interest‑driven feeds are essential, reviews Facebook's NewsFeed evolution and EdgeRank algorithm, details Pinterest's Smart Feed architecture and Pinnability model, and provides a comprehensive guide to building, ranking, and monitoring generic interest‑feed systems for social platforms.
This article introduces the concept and history of recommendation algorithms, outlines the basic conditions for recommendations, and provides a detailed explanation of user-based collaborative filtering, including similarity calculations, neighbor selection, recommendation scoring, practical code snippets, and discussion of potential issues.
This article introduces the main types of data‑mining algorithms—classification, prediction, clustering, and association—explains supervised and unsupervised learning, and illustrates each with practical examples such as spam detection, tumor identification, wine quality assessment, fraud detection, recommendation systems, and authorship analysis.
This article presents a comprehensive overview of ranking learning techniques used in Mobile Taobao's recommendation system, covering problem definition, pointwise/pairwise/listwise approaches, feature engineering, online learning, industry applications, and future optimization strategies.
The article explains how search engine techniques such as large‑scale candidate recall, fine‑grained ranking, user profiling, and multi‑objective optimization are applied to news personalization at Toutiao, highlighting data sampling, machine‑learning pipelines, challenges of news freshness, and architectural evolution.
This article outlines the four‑layer architecture of Weibo’s recommendation system—front‑end (RFront), application (RApp), compute (RCompute), and data (RStore)—detailing each layer’s purpose, technologies such as CRF, nginx+Lua, C++ frameworks, and supporting tools for logging, monitoring, and evaluation.
This talk shares the experience of a small machine‑learning team at Dangdang, describing how they built a recommendation system from scratch, the tools and processes they used, the challenges of limited personnel, and the many pitfalls they encountered while iterating toward a production‑ready solution.
Quora employs a variety of machine‑learning techniques—from ranking and personalized feed algorithms to duplicate‑question detection, user expertise inference, and content moderation—optimizing both user experience and content quality through offline testing, online A/B experiments, and models such as logistic regression, gradient‑boosted trees, and neural networks.
Meituan’s recommendation platform comprises a data layer, a multi‑strategy candidate generation layer, a fusion‑and‑filtering layer, and a ranking layer that uses additive‑grove tree ensembles and online‑updated logistic regression, leveraging extensive user behavior logs, location, query, graph and real‑time signals to deliver personalized deals.
