Efficient Short-Term Weather Forecasting with Random Forests: A Study on Limited Dataset

Accurate temperature forecasting plays a vital
role in supporting data-driven decision-making across various
sectors including energy planning, environmental comfort
management, and public safety. This study investigates the
application of Random Forest Regression in a limited dataset for
short-term air temperature prediction. Specifically, Random
Forest is designed to forecast hourly temperature values for the
next seven days using a minimal dataset upon a single day of
meteorological observations. The observation includes
temperature, humidity, atmospheric pressure, and wind speed.
Despite of the limited temporal scope of dataset, Random Forest
model demonstrates a notable ability to simulate realistic
temperature patterns. The results reveal that Random Forest is
capable of handling heterogeneous input features and delivering
accurate predictions under normal environmental condition
even with constrained data availability. Accordingly, this
research yields the convergence of MAE (Mean Absolute Error)
is 0.102 and RMSE (Root Mean Square Error) value is 0.136.
The findings underscore the potential of Random Forest for
short-term temperature forecasting in data-limited scenarios.