Data Reduction in Monitored Data

Nowadays, increasing time series data has brought new challenges in many domains because their massive instances, dimensions, speed and complexity. In order to solve this problem, data reduction techniques are becoming an integral part of future systems, especially for Monitoring System. Different from most data reduction techniques which are based on information theory, this paper is planning to explore a novel model-based method, which tries to build piecewise regression models only for correlated data with guidance of priori knowledge, avoiding unnecessary computation between unrelated data streams. Until now, we have designed a primal data reduction framework, corresponding experiments and evaluation criteria.