Error estimation and model consolidation for time series data

Deregulation in the electric power industry has resulted in short term electric power load forecasting becoming much more interesting. Large quantities of capital as well as social benefits are at risk in the act of supplying reliable electric power to the populous. Managing this risk requires planning and a knowledge of possible future events, hence, electric power load forecasting. The short term electric power load demand forecasting example provided demonstrates both the theoretical concepts and the application utility of modified series association (MSA) on an actual electric power demand prediction problem. The results show that MSA provides reliable error distribution estimates as well as improved models through consolidation in one step whereas stacked generalisation requires considerably more computation and manipulation.