考虑节假日的ARIMA模型在酸奶销量预测中的应用

Application of ARIMA Model Considering Holidays in Sales Forecasting

摘要: 商品的销量往往受多种因素的影响，进行销量预测时也需要将这些因素考虑进去。传统的ARIMA模型进行销量预测时只考虑了时序序列的线性因素，而忽略了一些非线性因素。本文以冷藏酸奶的日销量为研究对象，探究了加入节假日等非线性因素的ARIMA模型预测的精确度。具体地，本文首先提取了2021年冷藏酸奶日销量数据，并运用BIC值对ARIMA模型的参数进行选择，然后对加入节假日等非线性因素的ARIMA模型、普通ARIMA模型和目前流行的ARIMA + SVM组合模型进行了实验评估和对比。实验结果表明，相较于普通的ARIMA模型和目前流行的组合模型，加入非线性因素的ARIMA模型预测结果更加准确。该研究成果对于冷藏酸奶销量预测的实际应用和相关研究具有重要的参考意义。

Abstract: The sales volume of commodities is often affected by a variety of factors, and these factors need to be taken into account when conducting sales volume forecasting. The traditional ARIMA model for sales forecasting only considers the linear factors of time series and ignores some nonlinear factors. In this paper, we take the daily sales volume of refrigerated yogurt as the research object, and ex-plore the accuracy of the ARIMA model prediction by adding non-linear factors such as holidays. Specifically, this paper firstly extracts the data of daily sales of refrigerated yogurt in 2021 and ap-plies the BIC value to select the parameters of the ARIMA model, and then experimentally evaluates and compares the ARIMA model with nonlinear factors such as holidays, the ordinary ARIMA mod-el, and the combined ARIMA + SVM model. The experimental results show that compared with the ordinary ARIMA model and the current popular combination model, the ARIMA model with non-linear factors is more accurate in predicting results. The research results are of great reference sig-nificance for the practical application of refrigerated yogurt sales forecasting and related research.