Chemical Commodity Price Forecasting Based On Combinatorial Models

With the recovery of the national economy,the driving force of economic development penetrates into all walks of life,among which chemical commodities play an important role.As one of the basic industries of the national economy,chemical products are the cornerstone of agricultural development,providing important raw material guarantee for industrial and agricultural production,and their quality,quantity and price are closely paid by the upstream and downstream industries.In addition,chemical products are also the basis for national defense production supporting high-tech materials,which are closely related to people’s clothing,food,housing and transportation.Therefore,whether it is possible to accurately predict the price of various chemical products has strong practical and theoretical significance.Since the price of chemical products is affected by various factors,showing strong non-linearity,non-stationarity and no obvious trend,the subjective guessing method of artificial experience obviously cannot meet the demand,and the traditional single model is difficult to capture the hidden law of its data.Based on this,this paper proposes a combinatorial prediction model based on data decomposition,which can deeply explore the potential fluctuation characteristics of the data,so as to better grasp the fluctuation law and realize its price prediction.Taking "decomposition input and combining output" as the guiding ideology,this paper constructs a price prediction model based on multi-factor decomposition integration and a price prediction model based on data feature-driven modeling,which explains and predicts them from the factor level and the fluctuation law of the price series itself,respectively.The specific research content is as follows:(1)Price prediction based on multi-factor decomposition integration.In this paper,the stationarity test,unit root test and Granger causality test,are first used to filter and screen the possible influencing factors of chemical product prices,leaving the main driving factors.Secondly,the price series of chemical products is decomposed into several eigenmode functions by the EEMD algorithm,and the obtained eigenmode functions are recombined to form a new meaningful modal sequence by using the DTW reconstruction rule,which reduces the calculation amount of the model on the basis of ensuring the predictive ability of the model.Thirdly,the screened factors are substituted into the LSTM network to predict the reconstructed modal sequences,and the component prediction values of each mode are obtained,and finally the final prediction results are obtained through the integrated algorithm.(2)Decompose integrated price predictions based on data characteristics.In this section,factor sequences were removed as auxiliary support in prediction,and price prediction was mainly conducted from a time series perspective.Considering that the data characteristics displayed by each component modality of the price series after being decomposed by the EEMD algorithm are not completely consistent,mainly including periodicity,mutation,trendiness,and complexity,or a combination of multiple characteristics,combined with the adaptability of the prediction model,models are established separately by analyzing the data characteristics displayed by each modality,fully achieving "tailoring".Finally,in order to further reduce the computational complexity of the model and improve the prediction ability of the model,multiple modes that can be predicted by using the same model are introduced into the DTW reconstruction rule combination for prediction.Finally,the final prediction results are obtained by integrating the rules.Based on the above,this article has established a price prediction model based on influencing factors and a price prediction model based on time series fluctuations,explaining the prices of chemical products from two different levels.Through numerical empirical research,it has been shown that both models can effectively predict the prices of chemical products in their respective situations.