Модифікація методу ddm для вирішення задач потокої обробки даних з використанням сімейства моделей eTS

Abstract

In real-time machine learning systems, performance may degrade as environmental properties change. Therefore, when working with large data streams, there is a need to develop self-maintaining and self-learning systems. The purpose of incremental machine learning models is to solve this problem. Incremental models can learn and predict on-the-fly. Evolving fuzzy systems can operate in an online mode with the ability to adapt model architecture and parameter values. The evolving Takagi-Sugeno (eTS) family can provide accurate results for data streams. This work presents several methods to handle changes in the environment for eTS.