基于优化的长短时记忆神经网络牧群采食量估测模型

为了能够较为精准的估测牧群的采食量信息，提出一种基于遗传算法(genetic algorithm, GA)和长短时记忆神经网络（long short-term memory，LSTM）的牧群采食量估测模型。首先通过皮尔森系数法分析得出影响牧群的采食量的主要影响因子，以减少输入维度并解决信息冗余问题。在此基础上，构建基于 LSTM 神经网络算法的牧群采食量估测模型，并引入遗传算法来优化LSTM 神经网络模型参数来增加模型的可靠性。最后，利用该模型对牧群采食量进行估测。试验结果表明：该采食量估测模型各评价指标平均绝对误差(mean absolute error，MAE)、平均绝对百分比误差(mean absolute percentage error，MAPE)、以及均方根误差（root mean square error，RMSE）分别为2.982、9.85%和6.108。与单一的LSTM神经网络以及GRU神经网络模型相比，均优于其他模型；且该模型具有较好的估测性能和较强的泛化能力，能够为合理轮牧提供科学指导，对草地保护有一定的应用价值。

Estimation model of feed intake in herding sheep based on optimized long short-term memory neural network

In order to accurately estimate the feed intake information of herds, a herd feed intake estimation model based on genetic algorithm (GA) and long short-term memory neural network (LSTM) was proposed. Firstly, the main influencing factors affecting the feed intake of herds were analyzed by Pearson coefficient method to reduce the input dimension and solve the problem of information redundancy. On this basis, a herd feed intake estimation model based on LSTM neural network algorithm was constructed, and genetic algorithm was introduced to optimize the parameters of LSTM neural network model to increase the reliability of the model. Finally, the model was used to estimate the feed intake of herds. The results showed that the mean absolute error (MAE), mean absolute percentage error (MAPE) and root mean square error (RMSE) of the evaluation indexes of the feed intake estimation model were 2.982, 9.85% and 6.108, respectively. Compared with the single LSTM neural network and GRU neural network model, they were better than other models; the model had a good estimation performance and strong generalization ability, which can provide a scientific guidance for reasonable rotational grazing, and it has a certain application value for grassland protection.