Aprendiendo con ensambles a partir de flujos de datos no estacionarios.

Nowadays many sources generate massive data continuously, without control of the arrival order and at high speed. Internet, cell-phones, cars, and security sensors are examples of such sources. Because of the temporal dimension of the data (they are constantly arriving over time), and the dynamism of many real-world situations, the target function to be learned can change over time. This situation, known as concept drift, complicates the task of estimating this target function, because a previous learning model can become outdated, or even contradictory regarding the most recent data. There are several algorithms to manipulate concept drift, and among them are the classifier ensembles. In this article, we present a new ensemble algorithm called ADCE, able for learning from data streams with concept drift. ADCE manipulates these changes using a change detector in each base classifier. When the detector estimates a change, the classifier in which the change was estimated is replaced by a new one. ADCE combines the simplicity of the bagging algorithm to train base classifiers with methods used in batch learning to combine the output of the base classifiers. The proposed algorithm is compared empirically with several bagging family ensemble algorithms for data streams. The experimental results show that the proposed algorithm constitutes a viable option for learning from concept drifting data streams.