基于多维度综合决策的猪肉价格时间序列预测模型

打开文本图片集

Abstract：Shandong Province isamajorhub foranimal husbandry，andpork price fluctuations haveasignificantimpacton residents'qualityoflife.Currently，thereis limitedresearchonpredictinghogpricefluctuationsiShandongProvince，and issuessuchasshortpredictionhorizons，narowtimewindows，andinaccurateforecastingresultspersist.Toaddressthe insuficient acuracyof traditional prediction models in long-termtime series prediction，this paper proposes a time series predictionmodel based onacomprehensive decision-makingmechanism.Firstly，this study decomposes time series informationandamplifiesdatafeatures through reversiblenormalization to extract more pricefluctuation informationBased on the information decomposition， it expands prior knowledge via upsamplingand enhances thedata feature mining and decision-making capabilitiesof themulti-layer perceptron through multi-dimensional comprehensive decision-making. Finaly，itdirectlymaps prior knowledge topredictionresults，therebyaddressingtheisuesofnarrow windowsaderor accumulation caused bysliding window iterative prediction.Theexperimental results show thatcompared with models such as ARIMA，Prophet-BP、GA-BP，VMD-LSTM，and STL-Informer，thealgorithm inthispaperachievesanaverage improvement of 50.2% and 30.9% inRMSE and MAE indicators，respectively. Furthermore，it exhibits superior stability in the R2 indicators，with an average improvement of 60.2% over the aforementioned comparative models.The proposed algorithmexhibitsbetter forecasting performance forthe hog market in Shandong Province，which can assist relevant departments inmaking scientific decisionsregarding hog price fluctuations.

Keywords：Pork price;time series prediction; trend decomposition;up-sampling; multi-dimensional decision

中国是全球猪肉生产和消费大国，其生产和消费总量均位列世界第一，根据《中国统计年鉴》、美国农业部和经济合作与发展组织数据库的数据，2024年中国猪肉产量为5706万t，占全球猪肉总产量超过一半。（剩余16111字）