Scholarlink

This paper gives a summary of the Predictive Online Digital Sales (PODS) and Marketing Discovery Challenge, which took place during the ECML-PKDD conference in 2025. The PODS 2025 Challenge asked participants to forecast click-through rates (CTR) and conversion rates using data from real-time e-commerce campaigns. It was a challenging, practical task focused on predictive modeling in digital marketing. The challenge attracted 48 data scientists from around the world who submitted a total of 564 entries. The paper provides an overview of the problem, challenge setup, summary statistics on submissions, and brief descriptions of the winning solutions. It presents a comparative review of the top three winning solutions, along with a brief description of their adopted methodologies, a comparison of their respective performances, and a discussion of possible future directions. The contest was hosted on the Codabench platform ( Link to the challenge page ).

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predictive Online Digital Sales (PODS) and Marketing Challenge at the 2025 ECML-PKDD

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Anwesha Pal,
  • Hillol Kargupta,
  • Sanghamitra Bandyopadhyay,
  • Codrina Ana Maria Lauth,
  • Ernestina Menasalvas,
  • Katharina Morik

摘要

This paper gives a summary of the Predictive Online Digital Sales (PODS) and Marketing Discovery Challenge, which took place during the ECML-PKDD conference in 2025. The PODS 2025 Challenge asked participants to forecast click-through rates (CTR) and conversion rates using data from real-time e-commerce campaigns. It was a challenging, practical task focused on predictive modeling in digital marketing. The challenge attracted 48 data scientists from around the world who submitted a total of 564 entries. The paper provides an overview of the problem, challenge setup, summary statistics on submissions, and brief descriptions of the winning solutions. It presents a comparative review of the top three winning solutions, along with a brief description of their adopted methodologies, a comparison of their respective performances, and a discussion of possible future directions. The contest was hosted on the Codabench platform ( Link to the challenge page ).