Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{210094,
        author = {Roshini B and Manivannan B},
        title = {Buffer-Free Real-Time Forecasting under Concept Drift},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {5},
        pages = {4089-4094},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=210094},
        abstract = {Real-time forecasting systems in energy, transport, finance and industrial monitoring work on data streams whose statistical properties change over time, a problem known as concept drift. A forecasting model trained on past data loses accuracy when drift occurs, so it must notice the change and adapt. Existing drift-handling methods usually keep a sliding window, a replay buffer, a memory module or an ensemble of models to detect drift and update the forecaster. These designs increase memory use and delay, are hard to run on edge devices, and require raw samples to be stored. This paper surveys representative work on drift detection, drift-aware ensembles, input normalisation, online time series forecasting and online learning, compares their datasets, mechanisms and memory demands, and identifies open problems: dependence on windows, buffers or several models, slow reaction to gradual drift, the difficulty of choosing window sizes, and evaluation that rarely reports memory and latency. Based on the survey, we outline a conceptual buffer-free framework that normalises the stream with running statistics, forecasts with a single incremental model, detects drift from the prediction error and adapts the model using only the current observation. The framework is a design proposal. It has not been implemented or evaluated here, and no performance claims are made for it.},
        keywords = {Concept Drift, Real-Time Forecasting, Buffer-Free Learning, Online Learning, Data Streams, Drift Detection, Time Series.},
        month = {October},
        }