Accurately predicting values for dynamic data streams is a challenging task in decision and expert systems, due to high data flow rates, limited storage and a requirement to quickly adapt a model to new data. We propose an approach for correcting predictions for data streams which is based on a reliability estimate for individual regression predictions. In our work, we implement the proposed technique and test it on a real-world problem: prediction of the electricity load for a selected European geographical region. For predicting the electricity load values we implement two regression models: the neural network and the k nearest neighbors algorithm. The results show that our method performs better than the referential method (i.e. the Kalman filter), significantly improving the original streaming predictions to more accurate values. © 2011 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Bosnić, Z., Rodrigues, P. P., Kononenko, I., & Gama, J. (2011). Correcting streaming predictions of an electricity load forecast system using a prediction reliability estimate. Advances in Intelligent and Soft Computing, 103, 343–350. https://doi.org/10.1007/978-3-642-23169-8_37
Mendeley helps you to discover research relevant for your work.