RULE-BASED PREDICTION OF SHORT TERM ELECTRIC LOAD

Abstract

In this study we discuss the possibility to apply symbolic data mining methods to the problem of prediction. We employ our original algorithm KEX that is used for extraction of classification or prediction rules from data. When new data is coming, the active rules (rules with a fulfilled left-hand side) from the rule base are applied to the data and their weights are composed by the inference mechanism to the resulting weight of a given prediction. The presented approach is applied to the problem of short-term electric load forecasting.