基于改进主成分分析的LSTM水稻产量预测模型研究

Research on LSTM Rice Yield Prediction Model Based on Improved Principal Component Analysis

摘要: 作为农业大国，农业一直是我国经济发展的重要基石，而水稻作为我国主要的粮食作物，随着人口不断增长，其需求量也不断攀升，因此水稻的产量预测对我国农业的发展建设以及保障粮食安全具有重要意义。长短期记忆(LSTM)循环神经网络因其不仅能够较好地处理各因素间的非线性关系，且适合处理时间序列数据的预测问题，在作物产量预测领域应用前景良好。本文提出一种基于改进主成分分析(IPCA)的LSTM循环神经网络，对神经网络的输入进行数据降维，旨在提高神经网络训练的收敛速度，并消除输入数据间由于相关性导致的信息冗余，从而提高预测精度。

Abstract: As a major agricultural country, agriculture has always been an important cornerstone of our country’s economic development. As the main food crop in our country, with the continuous growth of the population, the demand for rice is also rising. Therefore, the prediction of rice production is very important for the development and construction of agriculture and guarantee of food security. Long short-term memory (LSTM) recurrent neural network has a good application prospect in the field of crop yield prediction because it can not only deal with the nonlinear relationship between various factors, but also is suitable for dealing with the prediction problem of time series data. In this paper, a LSTM recurrent neural network based on improved principal component analysis is proposed, which reduces the data dimension of the input of the neural network, aiming to improve the convergence speed of neural network and eliminate the information redundancy caused by the correlation between the input data, thereby improving prediction accuracy.