Application of Regularization to Ensemble of Classifiers for Drift Compensation in Electrical and Chemical Equipments

1 Department of Computer Science, Banasthali University, C Scheme, Sarojini Marg , Jaipur 302001 , Rajasthan, INDIA 2 Department of Computer Science and Engineering, Indian Institute of Technology, Banaras Hindu University, Varanasi 221005, Uttar Pradesh, INDIA ______________________________________________________________________________________ Abstract: Any electrical or chemical device undergoes a problem of concept drift with time due to various chemical, physical interactions of the environment elements with the exposed surface of the device and other factors such as aging and poisoning of the surface. This has been a major problem faced by such equipments when they are used as an experimental tool during any research and developments. In this paper, we introduce a hard machine learning approach to solve this problem by the application of regularization to the ensemble of classifiers for overcoming the time dependent drift occurrence. We have applied our regularized drift compensation algorithm on two real time data and one synthetic data. Our experiment finds best improvement on the application of Single Value Decomposition and Norm 2 combination regularization to the ensemble of classifiers. To the best of our knowledge, regularization has not yet been applied for drift correction in such electrical and chemical equipments.