Sand storm level prediction method based on Stacking integration strategy

Abstract

A sand storm level prediction method based on a Stacking integration strategy comprises the steps: taking a recurrent neural network R and a convolutional neural network C as a first-level classifier,inputting original weather sample data into the recurrent neural network R and the convolutional neural network C respectively, and obtaining corresponding first-level learning features; introducinga meta-classifier Q as a secondary classifier by utilizing a Stacking integration strategy, and combining the primary learning features to serve as the input of the secondary classifier; and taking the output of the secondary classifier as the finally predicted sand storm grade quantity. According to the invention, the time series data processing capability of the RNN and the high-dimensional feature extraction capability of the CNN are fused; the sand storm level prediction method has a wider prediction angle and better generalization ability, and the selection of a default activation function can improve the flexibility and generalization ability of the model, and a 1 * 1 convolution kernel replaces a full connection layer, and more features can be integrated, and better generalization performance is provided, and L2 regularization and Batch-are adopted; and the sand storm level prediction method has higher generalization performance. The generalization ability of the classifiers ofall levels and the prediction accuracy and precision of the whole classifier are improved through the Normalization or Dropout technology.

Description

technical field [0001] The invention belongs to the technical field of computers, in particular to a method for predicting sandstorm grades based on a Stacking integration strategy. Background technique [0002] As a natural disaster, sandstorms occur frequently in arid and semi-arid regions. As early as 70 million years ago, there were sandstorms on the earth. Since modern times, due to environmental reasons such as soil erosion, land desertification, and vegetation destruction, the number of sandstorms in northern my country, especially in the northwest region, has increased significantly, and sandstorms have an increasing impact on people's production and life. [0003] Traditional Meteorological Forecasting Weather forecasting is based on meteorological observation data, applying the principles and methods of synoptics, dynamic meteorology, and statistics to make qualitative or quantitative predictions of the weather conditions of a certain area or a certain place in the...

AI Technical Summary

Problems solved by technology

Due to the complexity of the causes of sandstorms and the huge amount of meteorological data, it is difficult for ordinary neural networks to fit or generalize

Images