A pseudo-inverse method as an alternative in forecasting geothermal energy consumption and palm fruit production

Abstract

Forecasting is one among important aspects in giving an overview of future event, and even in making a decision for the respective problems to the event. Sometimes forecasting is done by correlating two different events or objects. Technically this is done by comparing data sets from these events. The most common method used to see the correlation between two data sets is the regression method or linear regression when the relationship is assumed to be linear. In this paper we discuss a different method which use pseudo-inverse of a matrix resulting from the data sets of the events. As the examples, we use the method in forecasting world geothermal energy consumption during the period of 1990-2010 and the palm fruit production in Pt. Perkebunan (PERSERO) Medan palm plantation during the period of 2011-2012. We compare the result from the linear regression method and from the pseudo-inverse method. In obtaining the pseudo-inverse of the resulting matrix we use singular value decomposition (SVD) which implemented in MATLAB program. The results show that in these different examples the linear regression method outperform the pseudo-inverse method. In the case of geothermal energy consumption, the Mean Absolute Percentage Error (MAPE) are 0.02619 for the linear regression method and 0.22107 for the pseudo-inverse method. While in the case of palm fruit production, MAPE are 0.09913 for the linear regression method and 0.10369 for the pseudo-inverse method.