A novel discrete multivariable grey model with spatial proximity effects for economic output forecast

The temporal and spatial characteristics of the collected data typically play an essential role in time series forecasting. However, traditional grey models pay more attention to temporal information, meanwhile disregarding the presence of spatial features critical to producing accurate forecasts. Moreover, the grey theory has few attempts to overcome this problem. Given this situation, this paper proposes a novel multivariable grey prediction model considering the spatial proximity effect for time series forecasting. More specifically, the complete spatial distance index is defined initially based on the geographical and economic distance to quantify the intensity of the spatial proximity impact. On this basis, the spatial proximity effect term is constructed and incorporated into the conventional discrete multivariable grey model, establishing the novel model. Subsequently, the particle swarm optimization algorithm determines the optimal weight coefficient in the developed index. For illustration and verification purposes, the proposed model performs experiments predicting Jiangsu's economic outputs compared to five prevalent benchmarks. Furthermore, three statistical metrics, the Diebold-Mariano test, the Monte Carlo experiments, and the probability density analysis are employed to validate the model's efficacy and robustness. Empirical results demonstrate that the proposed method outperforms all five competitors with strong robustness and broad generalization. Therefore, this new model is used for forecasting Jiangsu's future economic outputs from 2021 to 2025.