Evolving Classifier TEDAClass for Big Data

Abstract In the era of big data, huge amounts of data are generated and updated every day, and their processing and analysis is an important challenge today. In order to tackle this challenge, it is necessary to develop specific techniques which can process large volume of data within limited run times. TEDA is a new systematic framework for data analytics, which is based on the typicality and eccentricity of the data. This framework is spatially-aware, non-frequentist and non-parametric. TEDA can be used for development of alternative machine learning methods, in this work, we will use it for classification (TEDAClass). Specifically, we present a TEDAClass based approach which can process huge amounts of data items using a novel parallelization technique. Using this parallelization, we make possible the scalability of TEDAClass. In that way, the proposed approach is particularly useful for various applications, as it opens the doors for high-performance big data processing, which could be particularly useful for healthcare, banking, scientific and many other purposes.