RecSys Challenge 2022:Fashion Prediction Competition

Keyword：Machine Learning, Recommendation system
RecSys Challenge 2022 was a data competition based on fashion clothing recommendations. Participants were required to predict the items that users would eventually purchase based on their browsing history sessions. Collaborative filtering algorithm is one of the most classic recommendation algorithms. For this challenge, I chose to model the high-order connection information between users and items using the Neural Graph Collaborative Filtering algorithm. By analyzing user behavior, the algorithm calculates the similarity between items and recommends the items that are most similar to the ones the user has browsed. The algorithm achieved an MMR (Mean Reciprocal Rank) of 0.18 on the test dataset, ranking 33rd among nearly 600 participants.