Dynamic integration of classifiers for handling concept drift

Abstract

Abstract. In the real world concepts are often not stable but change with time. A typical example of this in the biomedical context is antibiotic resistance, where pathogen sensitivity may change over time as new pathogen strains develop resistance to antibiotics that were previously effective. This problem, known as concept drift, complicates the task of learning a model from data and requires special approaches, different from commonly used techniques that treat arriving instances as equally important contributors to the final concept. The underlying data distribution may change as well, making previously built models useless. This is known as virtual concept drift. Both types of concept drifts make regular updates of the model necessary. Among the most popular and effective approaches to handle concept drift is ensemble learning, where a set of models built over different time periods is maintained and the best model is selected or the predictions of models are combined, usually according to their expertise level regarding the current concept. In this paper we propose the use of an ensemble integration technique that would help to better handle concept drift at an instance level. In dynamic integration of classifiers, each base classifier is given a weight proportional to its local accuracy with regard to the instance tested, and the best base classifier is selected, or the classifiers are integrated using weighted voting. Our experiments with synthetic data sets simulating abrupt and gradual concept drifts and with a real-world antibiotic resistance data set demonstrate that dynamic integration of classifiers built over small time intervals or fixed-sized data blocks can be significantly better than majority voting and weighted voting, which are currently the most commonly used integration techniques for handling concept drift with ensembles.

Generalized CMAC adaptive ensembles for concept-drifting data streams

“…One example in the real world is spam filtering, where only some particular types of spam may change with time, while the others could remain the same. In these cases, the accuracy of global models may fall, even if they still could be good experts in the stable parts of the data [2].…”

“…Apart from filtering for recommendation systems, concept drift is a central problem in many dynamically changing and non-stationary environments, including medicine [2], industry [3], education [4], and business [5]. In these dynamic scenarios the challenge is to process, in (near) real-time, large highspeed data streams in order to adapt the learning model by combining what has been learned in the old concept with the fresh information corresponding to the new concept.…”

Learning from streaming data with concept drift and imbalance: an overview

“…Tsymbal et al propose a strategy based on local concept drift [80,81]. They argue that many real-world scenarios of concept drift are in fact local phenomenon, relegated to a specific region of the feature space.…”

One-class classifiers with incremental learning and forgetting for data streams with concept drift

“…The last group consists of algorithms that incorporate a set of classifiers (Wang et al 2003;Stanley 2003;Tsymbal et al 2008). It has been shown that a collective decision can increase classification accuracy because the knowledge that is distributed among the classifiers may be more comprehensive.…”