{"title":"Fashion Recommendation with a real Recommender System Flow","authors":"Qi Zhang, Guohao Cai, Wei Guo, Yiqiu Han, Zhenhua Dong, Ruiming Tang, Liangbi Li","doi":"10.1145/3556702.3556792","DOIUrl":null,"url":null,"abstract":"In this technical report, we present our solution of RecSys Challenge 2022 focusing on the fashion recommendation. We produce recommendations in two steps: (i) the retrieval step, which generates a candidate item set based on multiple cheap-to-compute strategies; (ii) the ranking step: which rearranges the candidate items with a richer set of features. Specifically, we conduct various strategies to retrieve as many positive samples as possible in retrieval step and obtain the retrieval scores from these retrieval channels meanwhile. Then these scores along with some extracted features are involved in the ranking stage for modeling to generate the purchase prediction. In the final submission, we use six effective retrieval strategies in retrieval step and ensemble five ranking models by taking average of their outputs. Using our method, our team doubleQ achieved MRR 0.2013 on final test set which wins the 10 place, and the solution codes are available via https://github.com/doubleQ2018/recsys-challenge-2022.","PeriodicalId":141185,"journal":{"name":"Proceedings of the Recommender Systems Challenge 2022","volume":"96 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the Recommender Systems Challenge 2022","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3556702.3556792","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
In this technical report, we present our solution of RecSys Challenge 2022 focusing on the fashion recommendation. We produce recommendations in two steps: (i) the retrieval step, which generates a candidate item set based on multiple cheap-to-compute strategies; (ii) the ranking step: which rearranges the candidate items with a richer set of features. Specifically, we conduct various strategies to retrieve as many positive samples as possible in retrieval step and obtain the retrieval scores from these retrieval channels meanwhile. Then these scores along with some extracted features are involved in the ranking stage for modeling to generate the purchase prediction. In the final submission, we use six effective retrieval strategies in retrieval step and ensemble five ranking models by taking average of their outputs. Using our method, our team doubleQ achieved MRR 0.2013 on final test set which wins the 10 place, and the solution codes are available via https://github.com/doubleQ2018/recsys-challenge-2022.