Utility of Deep Learning Algorithms in Initial Flowering Period Prediction Models

The application of a deep learning algorithm (DL) can more accurately predict the initial flowering period Platycladus orientalis (L.) Franco. In this research, we applied DL to establish nationwide long-term prediction model P. and analyzed contribution rate meteorological factors via Shapely Additive Explanation (SHAP). Based on daily data major stations in China from 1963–2015 observation 23 phenological stations, established models by using recurrent neural network (RNN), long short-term memory (LSTM) gated unit (GRU). mean absolute error (MAE), percentage (MAPE), coefficient determination (R2) were used as training effect indicators evaluate accuracy. simulation results show that three are applicable China, with average accuracy GRU being highest, followed LSTM RNN, which is significantly higher than regression based accumulated air temperature. interpretability analysis, factor rates similar, 46 temperature type have highest 58.6% factors’ 0 5.48 × 10−4, stability related minimum has obvious fluctuations an standard deviation 8.57 10−3, might be plants sensitive low stress. change rule flowering, greater 98%, best, indicating potential flowering.