Chain Drugstore Sales Prediction Method Based on the Fusion of Self-Attention Mechanism and LightGBM

Abstract

Optimizing inventory management in the pharmaceutical industry relies significantly on accurate sales forecasting for chain drugstores. Sales predictions for these stores, however, are affected by data quality and temporal characteristics, which limit the effectiveness of traditional statistical, machine learning, and ensemble learning methods. To address these challenges, this study introduces a sales forecasting model called TS-LGBM, which utilizes a sliding window approach to preserve the sequential integrity of sales data and integrates neural networks with the Light Gradient Boosting Machine (LightGBM). By incorporating a self-attention mechanism into LightGBM, the TS-LGBM model aims to enhance predictive accuracy. The model’s efficacy is validated using the Rossman dataset from Kaggle, followed by a case study with actual data from Z-chain retail drugstores. This study further refines data by factoring in temporal characteristics of various drugs, the density of nearby drugstores within a specified radius, and regional attributes associated with each drugstore. To evaluate performance, five models—TS-LGBM, TS-XGB, LightGBM, XGBoost and LSTM—are compared experimentally. Findings indicate that TS-LGBM achieves superior prediction accuracy compared to the other models. This study is intended for practical applications, as accurate sales forecasts for chain drugstores can enhance supply chain management efficiency and reduce operational costs.

Keywords

Drugstore chain / sales forecasting / time series / self-attention / LightGBM

References

[1] AbbasimehrH, ShabaniM, YousefiM. An optimized model using LSTM network for demand forecasting. Computers & Industrial Engineering, 2020, 143: 106435

[2] ArL, SkP, AtS. A novel machine learning approach to predict sales of an item in e-commerce. 2022 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES), 2022Chennai, India, July 15–16, 2022

[3] BoxG E, JenkinsG M, ReinselG C, LjungG MTime Series Analysis: Forecasting and Control, 2015John Wiley & Sons

[4] BoxG E, PierceD A. Distribution of residual autocorrelations in autoregressive-integrated moving average time series models. Journal of the American Statistical Association, 1970, 65(332): 1509-1526

[5] CaglayanN, SatogluS I, KapukayaE N. Sales forecasting by artificial neural networks for the apparel retail chain stores. Intelligent and Fuzzy Techniques in Big Data Analytics and Decision Making: Proceedings of the INFUS 2019 Conference, 2020Istanbul, Turkey, July 23–25, 2019

[6] ChangX H, XiongA. Application of Adaboost-based random forest algorithm in medical sales forecasting. Computer Systems Applications, 2018, 27(2): 202-206

[7] ChenY, HeX, PeiZ, YiW, WangC, ZhangW, JiZ. Development of a time series e-commerce sales prediction method for short-shelf-life products using GRULightGBM. Applied Sciences, 2024, 14(2): 866

[8] ChenT, GuestrinC. Xgboost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD’16), 2016San Francisco, CA, USA, August 13–17, 2016

[9] CortesC. Support-Vector Networks. Machine Learning, 1995, 20(3): 273-297

[10] Di PilloG, LatorreV, LucidiS, ProcacciE. An application of support vector machines to sales forecasting under promotions. 4OR, 2016, 14: 309-325

[11] DuttaS R, DasS, ChatterjeeP. Smart sales prediction of pharmaceutical products. 2022 8th International Conference on Smart Structures and Systems (ICSSS), 2022Chennai, India, April 21–22, 2022

[12] ElmanJ L. Finding structure in time. Cognitive Science, 1990, 14(2): 179-211

[13] FauconnierG, TurnerM. Conceptual integration networks. Cognitive Science, 1998, 22(2): 133-187

[14] FengC, ChenZ D. Application of weighted combination model based on XGBoost and LSTM in sales forecasting. Computer Systems Applications, 2019, 10: 226-232

[15] FriedmanJ H. Greedy function approximation: A gradient boosting machine. Annals of Statistics, 2001, 29(5): 1189-1232

[16] GarnierR, BelletoileA. A multi-series framework for demand forecasts in E-commerce. arxiv Preprint arxiv, 2019

[17] KeG, MengQ, FinleyT, WangT, ChenW, MaW, LiuTY. LightGBM: A highly efficient gradient boosting decision tree. Advances in Neural Information Processing Systems (NIPS), 2017Long Beach, CA, USA, December 4–9, 2017

[18] HochreiterS, SchmidhuberJ. Long short-term memory. Neural Computation, 2016, 9: 1735-1780

[19] HuW, ZhangX. Commodity sales forecast based on ARIMA model residual optimization. 2020 5th International Conference on Communication, Image and Signal Processing (CCISP), 2020November 13–15, 2020, Chengdu, China

[20] IslamL, FarooquiM F, KhanA, WasiM, ShaikhT. Walmart sales analysis and prediction. International Journal of Advanced Research in Science. Communication and Technology, 2023, 3(2): 347-352

[21] JhaB K, PandeS. Time series forecasting model for supermarket sales using FB-Prophet. 2021 5th International Conference on Computing Methodologies and Communication (ICCMC), 2021Erode, India, April 8–10, 2021

[22] LiB, LvX, ChenJ. Demand and supply gap analysis of Chinese new energy vehicle charging infrastructure: Based on CNN-LSTM prediction model. Renewable Energy, 2024, 220: 119618

[23] LiuY, LanK, HuangF, CaoX, FengB, ZhuB. An Aggregate Store Sales Forecasting Framework based on ConvLSTM. In Proceedings of the 2021 5th International Conference on Compute and Data Analysis, 2021Sanya, CHina, February 2–4, 2021

[24] LuX TResearch on store sales prediction of ND company based on XGBoost method, 2019Beijing University of Technology

[25] LuoB J. Research on library management system based on hybrid clustering algorithm. Computer & Digital Engineering, 2018, 46(06): 1173-1177

[26] MassaroA, PanareseA, GiannoneD, GalianoA. Augmented data and XGBoost improvement for sales forecasting in the large-scale retail sector. Applied Sciences, 2021, 11(17): 7793

[27] MessaoudiF, LoukiliM, El GhaziM. Demand prediction using sequential deep learning model. 2023 International Conference on Information Technology (ICIT), 2023Amman, Jordan, August 9–10, 2023

[28] QiaoZ. Walmart sale forecasting model based on LightGBM. 2020 2nd International Conference on Machine Learning, Big Data and Business Intelligence (MLBDBI), 2020Taiyuan, China, October 23–25, 2020

[29] RamosP, SantosN, RebeloR. Performance of state space and ARIMA models for consumer retail sales forecasting. Robotics and Computer-Integrated Manufacturing, 2015, 34: 151-163

[30] ShiC YResearch on sales forecast of entity retail industry based on data mining, 2019Guangdong University of Technology

[31] SoltaninejadM, AghazadehR, ShaghaghiS, ZareiM. Using machine learning techniques to forecast Mehram company’s sales: A case study. Journal of Business and Management Studies, 2024, 6(2): 42-53

[32] TangX, GaoS, JiangZ. Ablending model combined DNN and LightGBM for forecasting the sales of airline tickets. Proceedings of the 2019 3rd International Conference on Computer Science and Artificial Intelligence(CSAI), 2019Normal, IL, USA, December 6–8, 2019

[33] TaylorS J, LethamB. Forecasting at scale. The American Statistician, 2018, 72(1): 37-45

[34] TingK M, WittenI H. Stacking bagged and dagged models. Proceedings of the 14th International Conference on Machine Learning (ICML’97), 1997Nashville, TN, USA, July 8–12, 1997

[35] Vallés-PérezI, Soria-OlivasE, Martínez-SoberM, Serrano-LópezA J, Gómez-SanchísJ, MateoF. Approaching sales forecasting using recurrent neural networks and transformers. Expert Systems with Applications, 2022, 201: 116993

[36] VaswaniA, ShazeerN, ParmarN, UszkoreitJ, JonesL, GomezA, PolosukhinI. Attention is all you need. Advances in Neural Information Processing Systems (NIPS), 2017Long Beach, CA, USA, December 4–9, 2017

[37] WangJ. A hybrid machine learning model for sales prediction. 2020 International Conference on Intelligent Computing and Human-Computer Interaction (ICHCI), 2020Sanya, China, December 4–6, 2020

[38] WangL H. Short-term sales prediction model for e-commerce based on BP-AdaBoost. Computer System Applications, 2021, 2: 260-264

[39] WeiH, ZengQ T. Research on sales forecast based on XGBoost-LSTM algorithm model. 2020 3rd International Symposium on Power Electronics and Control Engineering (ISPECE 2020), 2020Chongqing, China, November 27–29, 2020

[40] WengT Y, LiuW Y, XiaoJ. Supply chain sales forecasting based on lightGBM and LSTM combination model. Industrial Management & Data Systems, 2020, 120(2): 265-279

[41] WuJ JResearch onretail store sales predictionbased on machine learning, 2020Chang’an University

[42] Wu N, Green B, Ben X, O’Banion S (2020). Deep transformer models for time series forecasting: The influenza prevalence case. arxiv Preprint arxiv: 2001.08317.

[43] XieD R, ZhangS L. Machine learning model for sales forecasting by using XGBoost. 2021 IEEE International Conference on Consumer Electronics and Computer Engineering (ICCECE), 2021January 15–17, 2021, Guangzhou, China

[44] XuMResearch on sales data mining and prediction model based on deep learning, 2019Dalian University of Technology

[45] YanB, LiG H, LinR J. Hybrid sales forecasting model. Computer Engineering and Design, 2015, 36(3): 814-818

[46] YuX, QiZ, ZhaoY. Support vector regression for newspaper/magazine sales forecasting. Procedia Computer Science, 2013, 17: 1055-1062

[47] ZhangHVegetable sales prediction based on ensemble learning, 2021Yunnan Normal University

[48] ZhangL DResearch and application of spandex product sales forecasting technology, 2020University of Chinese Academy of Sciences (Shenyang Institute of Computing Technology, Chinese Academy of Sciences)

[49] ZhangY XResearch on apple sales prediction method based on deep learning, 2021Xi’an University of Architecture and Technology

[50] ZhuangY. Research on the application of mathematics in sales trend forecasting method. Marketing Circles, 2020, 24: 191-192

RIGHTS & PERMISSIONS

Systems Engineering Society of China and Springer-Verlag GmbH Germany

Just Accepted

This article has successfully passed peer review and final editorial review, and will soon enter typesetting, proofreading and other publishing processes. The currently displayed version is the accepted final manuscript. The officially published version will be updated with format, DOI and citation information upon launch. We recommend that you pay attention to subsequent journal notifications and preferentially cite the officially published version. Thank you for your support and cooperation.