基于分解重构LSTM的氟化工产品价格预测方法

A Method to Predict Fluoride Product Price Based on LSTM with Decomposition and Reconstruction

企业利润受产品价格波动影响,深入分析产品价格变化趋势可以显著提升企业的市场竞争力。由于传统产品销售价格分析无法有效预测市场变化,提出一种基于分解重构的长短期记忆(LSTM)方法,对氟化工产品价格进行分析和预测。该方法首先对特征数据进行预处理,包括缺失值、异常值和归一化处理;然后通过相关系数进行特征选择,并基于集成经验模态分解(EEMD)挖掘数据隐含信息,采用动态时间规整(DTW)算法进行聚类和重构,进而建立LSTM模型进行预测。实验表明,该模型可以提升预测精度、降低计算复杂度,且优于人工神经网络(ANN)和最小支持向量回归(LSSVR)等基准模型。

Corporate profits are affected by fluctuations in product prices. In-depth analysis of product price trends can significantly improve market competitiveness of enterprises. The traditional product pricing analysis cannot predict market changes effectively. An LSTM method based on decomposition and reconstruction is proposed in this paper to analyze and predict the price change of fluorine chemical products. First, the feature data are pre-processed, including missing values, outliers, and normalization. Then, perform feature selection is performed through correlation coefficients, and the hidden information of data is mined based on ensemble empirical mode decomposition(EEMD), dynamic time warping(DTW) algorithm is used for clustering and reconstruction. Finally, long short term memory model is established for prediction of the reconstructed data.The experimental results show that the LSTM model with decomposition and reconstruction proposed in this paper can improve the prediction accuracy and reduce the computational complexity. Besides, the model is better than benchmark models such as Artificial Neural Network(ANN) and Least Squares Support Vector Regression(LSSVR).