Forecasting Method Of Agricultural Futures Price Based On Multiple Influencing Factors

Among many financial products,agricultural product futures are the financial products which have the closest relationship with people's production and life.The price fluctuation of agricultural products futures will directly affect the spot price.Therefore,it is very important to construct a forecasting model of agricultural products futures price for market macro-control and risk prediction.In this paper,a forecasting model LSTM-FD for agricultural futures price based on multiple influencing factors is proposed,which can effectively improve the accuracy of price forecast by classifying the influencing factors.The main contents of this paper are as follows:First of all,this paper selects sugar futures as the research object,and puts forward seven factors that affect the price of sugar futures.This paper analyzes the causes of futures price fluctuation caused by influencing factors,extracts the characteristics of influencing factors and builds a mathematical model.Secondly,based on the impact of climate on white sugar crops,aiming at the asymmetry of FAD and the decrease of precision in dealing with long time series,FASM-WD,a time series clustering algorithm for climate data,is proposed in this paper.The algorithm can cluster and group the climate data and extract the characteristic values according to the climate characteristics required by crop growth.The experimental results show that FASM-WD algorithm improves the accuracy of climate time series data clustering compared with FASM algorithm.Then,in order to obtain the time series of futures data with different price trends,aiming at the problems of K-MA algorithm losing the characteristics of time series data,losing part of the data length,causing the time series peak and trough time shift,this paper proposes a time series denoising algorithm K-MDW.The algorithm combines the advantages of K-MA algorithm,DTW algorithm dynamic time warping time axis and wavelet transform dynamic window denoising advantages.It can construct different financial time series on the basis of time series noise reduction.Experimental results show that the denoising results of K-MDW algorithm are better than those of K-MA algorithm and wavelet transform.Finally,in order to solve the problem that different impact factors have different impacts on futures prices,the LSTM model can not deal with the input data of classified impact factors.In this paper,FASM-WD algorithm is used to obtain the characteristic value series of climate impact factors,and K-MDW algorithm is used to obtain different price trend series.The impact factors are one-to-one corresponding to different price trend series,and the agricultural product oriented period is proposed Price forecasting model LSTM-FD.The experimental results show that LSTM-FD model is more accurate than LSTM model,and the intermediate results of LSTM-FD model can be used for short-term,medium-term and long-term investment reference.