Główne pojęcia
Novel data stream regression method using database-inspired adaptive granulation for low-latency predictions.
Streszczenie
In the fast-paced world, time-sensitive systems require low-latency predictions from continuous data streams. Traditional regression techniques struggle with dynamic data, leading to the need for novel methods. The proposed model uses R*-trees for granulation, iteratively forgetting outdated information to maintain recent relevant granules. Experiments show significant latency improvement and competitive prediction accuracy compared to state-of-the-art algorithms. The approach is amenable to integration with database systems, offering scalability and efficiency.
Statystyki
Our method can be up to 6 times faster than ORTO.
Iterative Forgetting can be more than 10 times faster than ARF.
The model size of Iterative Forgetting consistently requires less space than other models.
Cytaty
"Our experiments demonstrate that the ability of this method to discard data produces a significant order-of-magnitude improvement in latency and training time."
"The R*-tree-inspired approach also makes the algorithm amenable to integration with database systems."
"Our model is developed with R* trees as a foundation, it can be implemented on the database level."