基于X12-ARIMA和LSTM组合模型的城市蔬菜价格波动规律及预测

Construction of urban vegetable price fluctuation prediction model based on X12-ARIMA and LSTM

蔬菜产业是社会的主导产业之一,是国民经济的重要组成部分,获取准确有效的蔬菜价格波动信息是种植技术员安排生产以及市场价格调控人员实施决策的重要依据.以成都市莴笋价格为例,提出一种基于时间序列的X12-ARIMA和长短期神经网络(LSTM)组合价格数据预测模型,重点比较研究组合模型的预测效果.首先通过H-P滤波方法和CensusX12对价格数据进行分解,分析其波动规律,再使用X12-ARIMA模型对蔬菜价格序列进行预测,揭示其线性特征,最后利用LSTM模型对X12-ARIMA预测模型的误差序列进行建模预测,得到蔬菜价格波动的非线性规律,二者预测值之和就是组合模型的预测值.实验结果表明,成都市蔬菜价格有季节性循环等特性,且通过对比和分析显示,提出的组合模型预测精度更高和性能更优,可作为预测城市蔬菜价格异常波动的参考依据.

The vegetable domain is one of the leading industries in society and an essential component of the country’s economy, and obtaining accurate and effective vegetable price fluctuation information is an important basis for planting technicians to arrange production, and for market price control personnel to implement decision-making. In this paper, the price of lettuce in Chengdu was taken as an example, and a combined price data prediction model based on X12-ARIMA and LSTM was proposed. The prediction effects of the combined model were compared and studied. Firstly, the price data were decomposed by H-P filtering method and Census X12 to analyze its fluctuation rule, then, the X12-ARIMA model was used for predicting the vegetable price series to reveal its linear feature, finally, LSTM model was used to model and predict the error series of X12-ARIMA prediction model, and the nonlinear law of vegetable price fluctuation was obtained, the sum of the predicted values of the two was the predicted value of the combination model. The results of the present study indicated that the vegetable price in Chengdu had the feature of seasonal cycle, and that through comparison and analysis, the combined model proposed in this paper had the best predictive ability, which can be used as a reference for predicting abnormal fluctuation of urban vegetable price.