Research Article
Understanding Elapsed-time Sampling Delayed Feedback
@INPROCEEDINGS{10.4108/eai.2-6-2023.2334607, author={Hanlang Zhao and Yiyun Quan and Hao Yu and Yongqi Wu and Zhengyuan Liu}, title={Understanding Elapsed-time Sampling Delayed Feedback}, proceedings={Proceedings of the 2nd International Conference on Information Economy, Data Modeling and Cloud Computing, ICIDC 2023, June 2--4, 2023, Nanchang, China}, publisher={EAI}, proceedings_a={ICIDC}, year={2023}, month={8}, keywords={delayed feedback es-dfm model}, doi={10.4108/eai.2-6-2023.2334607} }- Hanlang Zhao
Yiyun Quan
Hao Yu
Yongqi Wu
Zhengyuan Liu
Year: 2023
Understanding Elapsed-time Sampling Delayed Feedback
ICIDC
EAI
DOI: 10.4108/eai.2-6-2023.2334607
Abstract
ES-DFM is the model proposed recently to improve algorithmic performance in conversion rate in an E-commerce recommendation system. The model addresses the delayed feedback issue, which is a cutting-edge issue in terms of two core evaluation indices within a good recommendation system – user likeability and user behavior. Adding in data of false negatives, the model better weighed between the waiting time before updating the model training data and the freshness of the training data. In our research, we replicated the baselines – DFM, FNC, FNW, FSIW, ESDFM by trying on Google Co-lab and rented server. Results show that the codes run successfully. Then we experimented on the parameters of the ES-DFM model, in hope of optimizing the results even more. However, the change in parameters returned equally good performance but longer processing time.