 This paper proposes a new hybrid model for solar radiation prediction which combines the advantages of both frequency domain and sequence domain models. The frequency domain model uses wavelet transform, complete ensemble empirical mode decomposition, and multi-layer perceptron to predict solar radiation. This is followed by a light gradient boosting machine, LGBM, model to predict solar radiation in the sequence domain. Finally, a multi-layer perceptron-based hybrid model is constructed using the forecasts from the first two stages as inputs. The results show that the proposed hybrid model outperforms the frequency domain model by up to 36.38%. This article was authored by Jin Weng Park, Sung Woo Park, Jang Wah Shim, and others.