Research on Ensemble Prediction Model of Apple Flowering Date

Apple flowering freeze is one of the major disasters affecting yield, and prediction of flowering date using meteorological factors is one of the important aids to reduce the impact of freeze damage. The prediction results of one single model are prone to fluctuations due to interannual and spatial changes. In this paper, 6 areas were selected from Shandong, Shaanxi, Henan, and Liaoning. By analyzing the meteorological data and apple flowering date in the last 10 years, key meteorological factors affecting flowering date were selected based on distance correlation coefficients. Later, the ensemble prediction model was constructed by using support vector machine regression, multiple linear regression, and decision tree regression as the base models. The results showed that the mean absolute deviation of the ensemble prediction model was in the range of 0.736–3.616, which showed good stability and prediction accuracy and could provide theoretical support for apple flowering date prediction.