A short-term precipitation prediction method based on evolutionary neural network architecture search

Abstract

The application discloses a short-term precipitation prediction method based on evolutionary neural network architecture search, and belongs to the technical field of weather prediction. The application firstly constructs a coding rule of a neural network architecture, builds an initial short-term precipitation prediction model according to the coding rule, and takes the short-term precipitation prediction model as an individual in a population; then the individual in the population is iteratively optimized to obtain an optimal population in performance; the individual in the final population is evaluated in performance, and a final short-term precipitation prediction model is obtained according to the evaluation result, and the short-term precipitation prediction model is used to predict the precipitation condition. Compared with a short-term precipitation prediction model designed manually, the application can find an optimal neural network architecture suitable for a task more quickly for different meteorological data sets, and improve the accuracy and reliability of prediction.

Images