A2M Summit: AI & Machine Learning – Recommendation Algorithms in 58.com’s Industrial Transformation

A2M Summit: Artificial Intelligence & Machine Learning Summit

The A2M Summit aims to discover innovative projects and outstanding teams in AI, big data, and internet architecture, integrate cutting‑edge international technologies, build an industry case‑study think tank, and help Chinese enterprises successfully transform and upgrade in the AI era.

Conference Information Conference Name: A2M Summit – Artificial Intelligence & Machine Learning Summit Date: May 26‑27, 2023

Speaker Liu Lixi, Senior Algorithm Architect, Business Intelligence Department, 58.com – responsible for core algorithms and strategy mechanisms of the online advertising system, covering recommendation, search, creative, and policy optimization.

Talk Title Practice of Recommendation Algorithms in 58.com’s Industrial Transformation

Topic Overview: The presentation first introduces 58.com’s business characteristics and industrial transformation roadmap, then discusses recommendation algorithm practices in low‑frequency sparse scenarios, including feature interaction under data sparsity, behavior‑sequence learning for long‑term low‑frequency data, and multi‑scenario joint modeling. It further explores how to combine recommendation algorithms with industry traits, such as recommendation strategies under person‑job matching constraints and traffic distribution for supply‑demand scheduling. Finally, it examines ad‑creative optimization in recommendation scenarios, covering creative generation and selection.

Talk Outline: 1. 58.com’s business characteristics and industrial transformation roadmap 2. Recommendation algorithm practice in low‑frequency sparse scenarios 2.1 Feature interaction under data sparsity 2.2 Behavior‑sequence learning for low‑frequency long‑term data 2.3 Multi‑scenario joint modeling 3. Recommendation algorithm practice combined with industry traits 3.1 Recommendation strategies under person‑job matching constraints 3.2 Traffic distribution for supply‑demand scheduling 4. Ad‑creative optimization in recommendation scenarios 4.1 Creative generation (concatenation‑based + generative) 4.2 Creative selection (audience preference + diversity) 5. Summary

Key Highlights: 1. Solutions for recommendation in low‑frequency sparse scenarios 2. Integration of recommendation algorithms with industry characteristics 3. Ad‑creative optimization strategies within recommendation contexts

Audience Benefits 1. Understand applications and solutions for recommendation algorithms in low‑frequency sparse environments, improving effectiveness and accuracy. 2. Learn how to align recommendation strategies with industry constraints such as person‑job matching and supply‑demand scheduling. 3. Grasp the importance of ad‑creative optimization in recommendation systems to boost click‑through and conversion rates. 4. Gain practical experience and insights from real‑world deployments and future outlooks of recommendation technologies.