A New Method for Cigarette Sales Forecasting Using Empirical Mode Decomposition and Long Short-Term Memory Network

作者：T.Y.C. Wei, Yanbing Liu, Duoxi Xiao, Xue Song, Yan Chen · 年份：2024 · DOI：10.1145/3641181.3641190 · 被引用次数：2 · 研究领域：Advanced Chemical Sensor Technologies、Advanced Combustion Engine Technologies、Spectroscopy and Chemometric Analyses

Sales forecasting is crucial for effective marketing, inventory management, and production planning in the tobacco industry. However, predicting cigarette sales can be challenging due to the volatility of time series data, which is influenced by various factors. Traditional prediction methods often struggle to handle complex nonlinear relationships. To address this, this paper proposes a novel approach that combines empirical mode decomposition (EMD) and long-short-term memory networks (LSTM) to accurately forecast cigarette sales. The new method utilizes a rolling mechanism and comprises three main steps. Firstly, the EMD method is employed to decompose the original sales time series into several simple and relatively regular components. Next, LSTM is applied independently to predict each of these time series components. Finally, a straightforward addition and integration method is used to aggregate the individual prediction results, yielding the final sales prediction. To evaluate the performance of the proposed method, real sales data from tobacco companies is used and five long-term sales cigarette products are selected. The prediction results are then compared with those obtained from two benchmark models, LSTM and ARIMA. The empirical analysis demonstrates that the proposed model outperforms traditional prediction models in terms of both prediction accuracy and stability. These findings corroborate that the novel methodology provides a robust and reliable instrument for...