K-Regret Query Algorithm over Streaming Data
Yun Zhe An, Kai Chen, Rui Zhu · 2024
Selecting a portion of points that users are interested in many applications such as online search and recommendation when facing large amounts of data. The k-regret problem was recently proposed for the selection of high-quality objects. However, current research methods for k-regret queries are designed for static data environments and cannot effectively maintain query results in high-speed streaming data. In order to support this query, we proposes a query processing framework called KPSW (K-regret Partition Sliding Window) based on sliding windows. Firstly, KPSW utilizes the characteristics of data flow to partition windows. Secondly, it predicts the earliest time when new points entering the window become the predicted results, facilitate real-time provision of the latest results in high-speed flow environments. Finally, experimental results on several real and synthetic datasets show that KPSW runs faster than existing static algorithms, while providing almost identical quality result.