Prediction Of Price Index Of Fresh Waxy Corn In Northern Hebei Based On Time Series Reconstruction

Fresh glutinous corn is originally from China,with Wanquan District in northern Hebei province being the most representative region.Wanquan District has developed a fresh corn index based on data assets as a new production factor,but there has been relatively little research on regional glutinous corn price indices.This study selected fresh glutinous corn as the research object and collected planting,production,and sales data for the past three years by visiting farmers,processing companies,and agricultural departments.Based on data organization,processing,field experiments,and expert guidance,the fluctuation characteristics of the price index of fresh glutinous corn were analyzed from four aspects: trend,seasonality,periodicity,and irregularity.The main factors influencing the price index fluctuation were also analyzed from the perspectives of supply,demand,and abnormal situations.Traditional ARIMA and LSTM models were used to analyze and predict the price index,incorporating the former’s good predictive effect on linear parts and the latter’s characteristics of machine learning to construct an optimized ARIMA-LSTM combination model,which analyzed the relative situation of price level change trend and amplitude.Singular Spectrum Analysis(SSA)was also used to decompose the complex price index sequence into different component trend vectors with agreed proportions according to the window length.After machine judgment analysis,a few trend vectors with a higher contribution rate were selected to reconstruct the price index sequence,remove noise items,reduce sequence complexity,and reduce prediction interference,effectively simplifying the price index.The experimental results show that the experimental predicted values are closer to the real value trend,and it can achieve good predictive effects on the nonlinear part,which can better guide theoretical modeling and production practice.