关键词: 二氧化碳浓度预测, 多元数据变量间内在联系, 张量分解, 序列最小二乘规划, 长短 期记忆神经网络

Abstract: To address the problem that traditional prediction models cannot exploit the intrinsic connections among variables of multivariate data, this paper proposes a long and short-term memory (LSTM) neural network model, TD-LSTM-S, which is based on tensor decomposition and sequential least square quadratic programming (SLSQP) optimization. In the model, the data are constructed into tensors and are decomposed and optimized so that the data can retain the intrinsic connections among variables. The SLSQP algorithm is used to optimize the LSTM so that it can effectively use the intrinsic connections among variables to improve the prediction performance of the model. The experimental results show that the proposed TD-LSTM-S model has higher prediction performance than the traditional model.