Objective With the rapid development of big data technology and artificial intelligence, aiming at the problems of low accuracy, too large prediction area, too long model optimization time of the current rice yield prediction model, etc., a whale optimization algorithm-backpropagation (WOA-BP) rice yield prediction method based on Spark was proposed.
Method This paper took rice yield and weather data of counties/cities/districts in the western region of Guangdong Province as the research object, used WOA to optimize the weights and bias values of BP neural network, and constructed a rice yield prediction model to improve the prediction accuracy. In addition, the WOA-BP algorithm was parallelized in the Spark framework to reduce the algorithm time overhead.
Result In terms of model accuracy, by comparing the prediction results after inverse normalization, the mean absolute percentage error of the BP neural network model optimized by WOA decreased from 8.354% to 7.068%, and the mean absolute error decreased from 31.320 kg to 26.982 kg, the root mean square error dropped from 41.008 kg to 33.546 kg. In terms of run time, 3-node Spark cluster reduced runtime by 11 742 s over non-Spark mode, reducing time overhead by 44%.
Conclusion The WOA-BP rice yield prediction method based on Spark can better predict rice yield in western Guangdong counties/cities/districts, and at the same time can well reflect the influence of weather factors on rice yield in western Guangdong Province, which is a reference for studying the rice yield situation in western Guangdong counties/cities/districts and even the whole Guangdong.
DownLoad: