LSTM model prediction results.

The vegetable sector is a vital pillar of society and an indispensable part of the national economic structure. As a significant segment of the agricultural market, accurately forecasting vegetable prices holds significant importance. Vegetable market pricing is subject to a myriad of complex influences, resulting in nonlinear patterns that conventional time series methodologies often struggle to decode. In this paper, we exploit the average daily price data of six distinct types of vegetables sourced from seven key wholesale markets in Beijing, spanning from 2009 to 2023. Upon training an LSTM model, we discovered that it exhibited exceptional performance on the test dataset. Demonstrating robust predictive performance across various vegetable categories, the LSTM model shows commendable generalization abilities. Moreover, LSTM model has a higher accuracy compared to several machine learning methods, including CNN-based time series forecasting approaches. With R2 score of 0.958 and MAE of 0.143, our LSTM model registers an enhancement of over 5% in forecast accuracy relative to conventional machine learning counterparts. Therefore, by predicting vegetable prices for the upcoming week, we envision this LSTM model application in real-world settings to aid growers, consumers, and policymakers in facilitating informed decision-making. The insights derived from this forecasting research could augment market transparency and optimize supply chain management. Furthermore, it contributes to the market stability and the balance of supply and demand, offering a valuable reference for the sustainable development of the vegetable industry.

蔬菜产业是社会发展的重要支柱，亦是国民经济结构中不可或缺的组成部分。作为农产品市场的重要组成板块，精准预测蔬菜价格具有重大现实意义。蔬菜市场定价受诸多复杂因素影响，呈现出传统时间序列方法往往难以解析的非线性特征。本研究采集了2009年至2023年间，北京七大核心批发市场的六种不同品类蔬菜的日均价格数据。在训练长短期记忆网络（LSTM）模型后，我们发现其在测试集上展现出了卓越的预测性能。该LSTM模型在各类蔬菜品类上均表现出稳健的预测性能，展现出优异的泛化能力。此外，相较于包括基于卷积神经网络（CNN）的时间序列预测方法在内的多种机器学习方法，LSTM模型的预测精度更高。本研究的LSTM模型决定系数（R²）达0.958，平均绝对误差（MAE）为0.143，相较于传统机器学习基准模型，预测精度提升超过5%。因此，通过预测未来一周的蔬菜价格，我们期望该LSTM模型能够落地应用，助力种植者、消费者及政策制定者做出科学决策。本预测研究所得出的结论能够提升市场透明度，优化供应链管理。此外，该研究还有助于维持市场稳定、平衡供需关系，为蔬菜产业的可持续发展提供重要参考依据。