Substitutes and Complementary Products Modeling for Assortment Optimization

Jedna z 10 najlepszych prelekcji Data Science Summit: Machine Learning Edition 2021!

Sergiy Tkachuk & Szymon Łukasik - Substitutes and Complementary Products Modeling for Assortment Optimization - A Machine Learning Approach

Identification of substitutionary and complementary links between products in large-scale retail is a fundamental problem leading to assortment optimization issues. Substitutable products are interchangeable and can be purchased instead of each other, e.g., one mineral water for another. Complementary products might be bought and used together; they experience joint demand, e.g., laptops and mice. Each consumer has her preferences, making it extremely hard to predict the links between particular products in broad assortments, given various constraints and potential computational overhead. Besides discussing how this problem can be formulated in the domain of machine learning, we will also demonstrate our way of solving it - using Synerise’s Cleora graph embeddings. The presentation of the method will be followed by the results of its experimental evaluation and comparison with the other, state-of-the-art algorithm.

Poziom trudności: średnio-zaawansowany
Język: angielski