6. CONCLUSION
6.4 Proposals for future research
This study reveals potential directions for future research. To begin with, the models designed in this study could be developed further. One could aim to eliminate the limita-tions presented in the previous chapter by 1) dividing the products into classes based on different criteria, 2) testing a wider range of candidate sets when tuning the hyperparam-eters, 3) designing a separate feature engineering process for each model, and 4) using a data set with different time frame and aggregation. At least the selected model could be critically examined and further developed to make sure that the assumptions and decisions behind its design are optimal. In addition, since only five machine learning models were tested in this study, it would be interesting to test more machine learning models or completely different forecasting methods.
To generalize the results of this study, the machine learning models could be applied using data from other retail stores. Moreover, the classification approach could be stud-ied in other use cases than demand forecasting. As Bacchetti and Saccani (2012) state, classification of stock keeping units has not received much academic attention, even though it could be a powerful tool for identifying not only demand structures but also other characteristics, such as costs and supply uncertainty. In this study, new products were successfully classified to forecast demand, so it would be tempting to see if other product characteristics, for example price elasticity or contribution to the sales of the whole company, could be identified with classification methods.
In this study, it was recognized that regression algorithms were unsuccessful to forecast new product demand using the data of the case company, but it’s not known why that happened. One potential reason could be the limited amount of data. As a consequence, regression algorithms could be examined further using a larger data set. Currently, it’s difficult to increase the amount of data in the context of the case company. However, the amount of available data is continuously increasing with accelerating speed as the case company is growing rapidly. Hence, machine learning models based on regression algo-rithms may come into question in the case company in the future. Naturally, regression algorithms could be tested using larger data sets available from other sources already today.
Finally, larger amount of data would make it possible to successfully design separate models for different product types. For example, the case company sells both food and
non-food products, and these two product types might have fundamentally different be-haviour patterns. Even though the used feature, vat rate, distinguishes the food and non-food samples from each other, it’s not known how well one common model can identify the possible differences. Hence, it would be interesting to examine the model perfor-mance if separate models were designed for each product type.
REFERENCES
Bacchetti, A., & Saccani, N. (2012). Spare parts classification and demand forecasting for stock control: Investigating the gap between research and practice. Omega (Oxford), 40(6), 722–
737. https://doi.org/10.1016/j.omega.2011.06.008.
Bajari, P., Nekipelov, D., Ryan, S. P., & Yang, M. (2015). Machine Learning Methods for Demand Estimation. The American Economic Review, 105(5), 481–485.
https://doi.org/10.1257/aer.p20151021.
Batina, L., Gierlichs, B., Prouff, E., Rivain, M., Standaert, F. X., & Veyrat-Charvillon, N. (2011).
Mutual Information Analysis: a comprehensive study. Journal of Cryptology, 24(2), 269–291.
https://doi.org/10.1007/s00145-010-9084-8.
Buda, M., Maki, A., & Mazurowski, M. A. (2018). A systematic study of the class imbalance prob-lem in convolutional neural networks. Neural Networks, 106, 249–259.
https://doi.org/10.1016/j.neunet.2018.07.011.
Carbonneau, R., Laframboise, K., & Vahidov, R. (2008). Application of machine learning tech-niques for supply chain demand forecasting. European Journal of Operational Research, 184(3), 1140–1154. https://doi.org/10.1016/j.ejor.2006.12.004.
Chandramouli, S. (2018). Machine Learning (S. Dutt & A. Das, Eds.; 1st edition). Pearson Edu-cation India.
Chen, T., & Guestrin, C. (2016). XGBoost. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 13-17-, 785–794.
https://doi.org/10.1145/2939672.2939785.
Chen, X., Pang, Z., & Pan, L. (2014). Coordinating Inventory Control and Pricing Strategies for Perishable Products. Operations Research, 62(2). https://doi.org/10.1287/opre.2014.1261.
Dahouda, M. K., & Joe, I. (2021). A Deep-Learned Embedding Technique for Categorical Fea-tures Encoding. IEEE Access, 9, 114381–114391. https://doi.org/10.1109/AC-CESS.2021.3104357.
Ertugrul, O. F. (2018). A novel type of activation function in artificial neural networks: Trained
activation function. Neural Networks, 99, 148–157.
https://doi.org/10.1016/j.neunet.2018.01.007.
Evangelopoulos, N. E. (2013). Latent semantic analysis. Wiley Interdisciplinary Reviews. Cogni-tive Science, 4(6), 683–692. https://doi.org/10.1002/wcs.1254.
Fildes, R., Ma, S., & Kolassa, S. (2019). Retail forecasting: Research and practice. International Journal of Forecasting. https://doi.org/10.1016/j.ijforecast.2019.06.004.
Fisher, M. L., Hammond, J. H., Obermeyer, W. R., & Raman, A. (1994). Making supply meet demand: in an uncertain world. Harvard Business Review, 72(3), 83–94.
Galli, S. (2020). Python feature engineering cookbook : over 70 recipes for creating, engineering, and transforming features to build machine learning models (1st edition). Packt Publishing Ltd.
Hancock, J. T., & Khoshgoftaar, T. M. (2020). Survey on categorical data for neural networks.
Journal of Big Data, 7(1), 1–41. https://doi.org/10.1186/s40537-020-00305-w.
Hastie, T., Tibshirani, R., & Friedman, J. (2009). Elements of Statistical Learning. Springer.
Heale, R., & Twycross, A. (2015). Validity and reliability in quantitative studies. Evidence-Based Nursing, 18(3), 66–67. https://doi.org/10.1136/eb-2015-102129.
Holweg, M., Disney, S., Holmström, J., & Småros, J. (2005). Supply chain collaboration: Making sense of the strategy continuum. European Management Journal, 23(2), 170–181.
https://doi.org/10.1016/j.emj.2005.02.008.
Huber, J., & Stuckenschmidt, H. (2020). Daily retail demand forecasting using machine learning with emphasis on calendric special days. International Journal of Forecasting, 36(4).
https://doi.org/10.1016/j.ijforecast.2020.02.005.
Jordan, M. I., & Mitchell, T. M. (2015). Machine learning: Trends, perspectives, and prospects.
Science (American Association for the Advancement of Science), 349(6245), 255–260.
https://doi.org/10.1126/science.aaa8415.
Kaspi, O., Girshevitz, O., & Senderowitz, H. (2021). PIXE based, Machine-Learning (PIXEL) sup-ported workflow for glass fragments classification. Talanta (Oxford), 234, 122608–122608.
https://doi.org/10.1016/j.talanta.2021.122608.
Khalid, S., Khalil, T., & Nasreen, S. (2014). A survey of feature selection and feature extraction techniques in machine learning. 2014 Science and Information Conference, 372–378.
https://doi.org/10.1109/SAI.2014.6918213.
Kotsiantis, S. B. (2007). Supervised machine learning: A review of classification techniques.
Informatica (Ljubljana), 31(3), 249–268.
Kotsiantis, S. B., Kanellopoulos, D., & Pintelas, P. E. (2006). Data preprocessing for supervised leaning. International Journal of Computer Science, 1.
Kuhn, M., & Johnson, K. (2013). Applied Predictive Modeling. Springer New York.
https://doi.org/10.1007/978-1-4614-6849-3.
Kureshi, S., & Thomas, S. (2019). Online grocery retailing – exploring local grocers beliefs. Inter-national Journal of Retail & Distribution Management, 47(2). https://doi.org/10.1108/IJRDM-05-2018-0087.
Lewis, C. D. (Colin D. (1997). Demand forecasting and inventory control : a computer aided learn-ing approach. Woodhead Pub. Ltd. in association with the Institute of Operations Manage-ment. https://doi.org/10.4324/9781856179898.
Li, J., Cheng, K., Wang, S., Morstatter, F., Trevino, R., Tang, J., & Liu, H. (2018). Feature Selec-tion. ACM Computing Surveys, 50(6), 1–45. https://doi.org/10.1145/3136625.
Liu, Y., Zhou, Y., Wen, S., & Tang, C. (2014). A Strategy on Selecting Performance Metrics for Classifier Evaluation. International Journal of Mobile Computing and Multimedia Communi-cations, 6(4), 20–35. https://doi.org/10.4018/IJMCMC.2014100102.
Mahmoud, E. (1984). Accuracy in forecasting: A survey. Journal of Forecasting, 3(2), 139–159.
https://doi.org/10.1002/for.3980030203.
Manco, L., Maffei, N., Strolin, S., Vichi, S., Bottazzi, L., & Strigari, L. (2021). Basic of machine learning and deep learning in imaging for medical physicists. Physica Medica, 83, 194–205.
https://doi.org/10.1016/j.ejmp.2021.03.026..
Mentzer, J. T., & Cox JR, J. E. (1984). Familiarity, application, and performance of sales
forecast-ing techniques. Journal of Forecasting, 3(1), 27–36.
https://doi.org/10.1002/for.3980030104.
Quemy, A. (2020). Two-stage optimization for machine learning workflow. Information Systems (Oxford), 92, 101483. https://doi.org/10.1016/j.is.2019.101483.
Rana, S. M. S. (2020). Supply chain drivers and retail supply chain responsiveness: strategy as moderator. International Journal of Management Practice, 13(1).
https://doi.org/10.1504/IJMP.2020.104066.
Renko, S., & Ficko, D. (2010). New logistics technologies in improving customer value in retailing service. Journal of Retailing and Consumer Services, 17(3).
https://doi.org/10.1016/j.jretconser.2010.03.012.
Saunders, M. N. K. (2019). Research methods for business students (P. Lewis & A. Thornhill, Eds.; Eighth edition.). Pearson Education.
Shalev-Shwartz, S., & Ben-David, S. (2014). Understanding machine learning: From theory to algorithms. Cambridge University Press.
Shankar, V., Kalyanam, K., Setia, P., Golmohammadi, A., Tirunillai, S., Douglass, T., Hennessey, J., Bull, J. S., & Waddoups, R. (2021). How Technology is Changing Retail. Journal of Re-tailing, 97(1). https://doi.org/10.1016/j.jretai.2020.10.006.
Syntetos, A. A., Babai, Z., Boylan, J. E., Kolassa, S., & Nikolopoulos, K. (2016). Supply chain forecasting: Theory, practice, their gap and the future. European Journal of Operational Re-search, 252(1), 1–26. https://doi.org/10.1016/j.ejor.2015.11.010.
Wan, S., Liang, Y., Zhang, Y., & Guizani, M. (2018). Deep Multi-Layer Perceptron Classifier for Behavior Analysis to Estimate Parkinson’s Disease Severity Using Smartphones. IEEE Ac-cess, 6, 36825–36833. https://doi.org/10.1109/ACCESS.2018.2851382.
Williamson, K. C., Spitzer, D. M., & Bloomberg, D. J. (1990). Modern Logistics Systems: Theory and Practice. Journal of Business Logistics, 11(2), 65.
Yang, L., & Shami, A. (2020). On hyperparameter optimization of machine learning algorithms:
Theory and practice. Neurocomputing (Amsterdam), 415, 295–316.
https://doi.org/10.1016/j.neucom.2020.07.061.
Zentes, J., Morschett, D., & Schramm-Klein, H. (2012). Strategic Retail Management. Gabler Ver-lag. https://doi.org/10.1007/978-3-8349-6740-4.
Zheng, A. (2018). Feature engineering for machine learning: principles and techniques for data scientists (A. Casari, Ed.; First edition.). O’Reilly.
APPENDIX A: EVALUATION METRICS FOR THE MODELS
Data setModel avgminmaxavgminmaxavgminmaxavgminmax Support vector classification (min-max scaled features)0.6700.6700.6700.6690.6690.6690.6700.6700.6700.6690.6690.669 Nearest neighbors (min-max scaled features)0.6480.6480.6480.6500.6500.6500.6480.6480.6480.6490.6490.649 XGBoost (features not scaled)0.7350.7350.7350.7350.7350.7350.7350.7350.7350.7340.7340.734 Logistic regression (features not scaled)0.6860.6860.6860.6830.6830.6830.6860.6860.6860.6840.6840.684 Multi-layer perceptron (min-max scaled features)0.6740.6560.6910.6800.6660.6910.6850.6620.7060.6790.6510.705 Support vector classification (min-max scaled features)0.9550.9550.9550.9550.9550.9550.9550.9550.9550.9550.9550.955 Nearest neighbors (min-max scaled features)1.0001.0001.0001.0001.0001.0001.0001.0001.0001.0001.0001.000 XGBoost (features not scaled)1.0001.0001.0001.0001.0001.0001.0001.0001.0001.0001.0001.000 Logistic regression (features not scaled)0.7010.7010.7010.6980.6980.6980.7010.7010.7010.6990.6990.699 Multi-layer perceptron (min-max scaled features)0.6850.6620.7060.6900.6810.7040.6850.6620.7060.6790.6510.705 Tes
t in Tra
F1-scoreRecallPrecisionAccuracy