Comparison Between Elementary Network Structure and Prophet Model for Time-Series Forecasting: Merchandise Sales

Xu Zhang; Lixian Zhang; Ziang Yang

doi:10.54097/21h9k153

Authors

Xu Zhang
Lixian Zhang
Ziang Yang

DOI:

https://doi.org/10.54097/21h9k153

Keywords:

BiLSTM, Transformer, Prophet, RMSE, MAPE.

Abstract

In order to maximize the profits of supermarkets, this article discusses different models in relation to time series prediction, based on previous time points, to help with setting the merchandise price. In detailed, we compared 2 deep learning methods(BiLSTM and Transformer) and Prophet model for predicting the future sales information. The overall dataset comes from certain information in the cost unit price of each category in a certain market within continuous 61 days. We evaluated these models based on the gaps bewteen their predictions and actual values, as well as the model’s Interpretability. Our results indicate that the BiLSTM model could predict smoothly, while the Transformer excels in short-term prediction accuracy. The Prophet model, though less precise, offers valuable insights into seasonal patterns. We find that no single model is universally superior; instead, their efficacy varies with the nature of the sales data. A combined approach, utilizing the strengths of each model, may provide the most comprehensive forecasting strategy. This study underscores the importance of model selection in enhancing retail pricing strategies through accurate sales forecasting.

Downloads

Download data is not yet available.

References

Dama F, Sinoquet C. Time series analysis and modeling to forecast: A survey[J]. arXiv preprint arXiv:2104.00164, 2021.

Box G E P, Jenkins G M, Reinsel G C, et al. Time series analysis: forecasting and control[M]. John Wiley & Sons, 2015.

Rai N K, Saravanan D, Kumar L, et al. RMSE and MAPE analysis for short-term solar irradiance, solar energy, and load forecasting using a Recurrent Artificial Neural Network[M]//Applications of AI and IOT in Renewable Energy. Academic Press, 2022: 181-192.

Schuster M, Paliwal K K. Bidirectional recurrent neural networks[J]. IEEE transactions on Signal Processing, 1997, 45(11): 2673-2681.

S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Comput., vol. 9, no. 8, pp. 1735_1780, 1997.

Siami-Namini S, Tavakoli N, Namin A S. The performance of LSTM and BiLSTM in forecasting time series[C]//2019 IEEE International conference on big data (Big Data). IEEE, 2019: 3285-3292.

Zeroual A, Harrou F, Dairi A, et al. Deep learning methods for forecasting COVID-19 time-Series data: A Comparative study[J]. Chaos, solitons & fractals, 2020, 140: 110121.

Liu Z-x , Zhang D-g , Luo G-z , Lian M , Liu B . A new method of emotional analysis based on CNN–BiLSTM hybrid neural network. Cluster Comput 2020:1–13 .

Graves A , Jaitly N , Mohamed A-r . Hybrid speech recognition with deep bidi-rectional LSTM. In: 2013 IEEE workshop on automatic speech recognition and understanding. IEEE; 2013. p. 273–8 .

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is All You Need. In Advances in Neural Information Processing Systems (pp. 5998-6008)

Taylor S J, Letham B. Forecasting at scale[J]. The American Statistician, 2018, 72(1): 37-45.

Zeng Y-R, Zeng Y, Choi B, Wang L (2017b) Multifactor-influenced energy consumption forecasting using enhanced back-propagation neural network. Energy 127:381–396.

Liu B, Song C, Wang Q, et al. Forecasting of China’s solar PV industry installed capacity and analyzing of employment effect: based on GRA-BiLSTM model[J]. Environmental Science and Pollution Research, 2022, 29(3): 4557-4573.