Change Detection with Kalman Filter and CUSUM

Knowledge discovery systems are constrained by three main limited resources: time, memory and sample size. Sample size is traditionally the dominant limitation, but in many present-day data-mining applications the time and memory are the major limitations [6]. Several incremental learning algorithms have been proposed to deal with this limitations (e.g., [5, 12, 6]). However most learning algorithms, including the incremental, make the assumption that the examples are draw from stationary distribution [13]. The aim of this study is to present a detection system (DSKC) for regression problems. The system is modular and works as a post-processor of a regressor. It is composed by a regression predictor, a Kalman filter and a Cumulative Sum of Recursive Residual (CUSUM) change detector. The system continuously monitors the error of the regression model. A significant increase of the error is interpreted as a change in the distribution that generates the examples over time. When a change is...

• Type: Conference
• Year: 2006
• Where: DIS
• Authors: Milton Severo, João Gama
• Keywords: DIS 2006, Learning Algorithms, Regression Model, Sample Size, Theoretical Computer Science