Forecasting the Consumption for Electricity in Taiwan

This paper use linear regression and non-linear artificial neural network (ANN) model to analyze how the four economic factors: national income (NI), population (POP), gross of domestic production (GDP), and consumer price index (CPI), affect Taiwan’s electricity consumption, furthermore, develop an economic forecasting model. Both models agree with that POP and NI are of the most influence on electricity consumption, whereas GDP of the least. Then, we compare the out-of-sample forecasting capabilities of the two models. The comparing result indicates that the linear model is obviously of higher bias value than that of ANN model, and of weaker ability of forecasting capability on peaks or bottoms. This probably results from: 1) linear regression model is built on the logarithm function of electricity consumption, and ANN is built on the original data; 2) ANN model is capable of catching sophisticated non-linear integrating effects. Consequently, ANN model is the more appropriate between the two to be applied to building an economic forecasting model of Taiwan’s electricity consumption.