Temperature forecasting using Dynamic Mode Decomposition

Variation of temperature in an area can have a high influence on the given area and therefore forecasting of temperature can have various benefits. This paper aims to forecast temperature using Dynamic Mode Decomposition (DMD). DMD is a completely data-driven algorithm. The DMD modes that capture the dynamic behavior of the system can be used for predicting the future behavior of the system. Temperature prediction can be done using various statistical techniques or machine learning techniques like the ARIMA model, RNNs, etc. Unlike machine learning models that are presently used, DMD does not require any training. There is not much literature exploring the use of DMD for this purpose. This work has taken daily temperature data and predicted the temperature for the next few days. Different sampling window was used for the prediction. The Root Mean Squared Error was used to calculate the error.