Probabilistic Ambient Temperature Forecasting Using Quantile Regression Averaging Model

Probabilistic ambient temperature forecasting is fascinating research in recent times. Rejuvenating the existing models with hybridization to improve forecast accuracy has a broad research interest. The nonparametric forecast combination technique, such as quantile regression averaging, is powerful enough with various scopes for improvements. The choice of different number and types of point forecasters as regressors is studied here. Two similar quantile regression averaging approaches are compared with a diverse selection of individual models, and the use of theoretically proposed regressors is the main highlight of the study. They are compared by using simulations with real-world ambient temperature data collected from different places in India. The proposed quantile regression approach is potentially better than the one that doesn't use the theoretical regressors.