The efficiency of data mining in Supply Chain Management (SCM) is becoming increasingly prominent and crucial for business success. SCM involves coordinating production, distribution, and transportation processes both within and outside the company to improve its performance (Sener et al. 2019). Generally, these activities entail a considerable amount of data, especially in transnational corporations. However, over recent years, approaches to data mining have made an innovative leap and are being actively implemented by the business. According to Wang et al. (2016), supply chain analysis methods vary depending on the objectives, which include descriptive, predictive, and prescriptive. This paper describes major data mining approaches, defines descriptive data mining techniques used in supply chain analysis, and discusses areas of logistics and SCM where specific approaches and techniques are used.
Data Mining Approaches
The function and purpose of the data mining vary depending on the area of application. Data analysts are facing three global challenges in the logistics field: the volume, asymmetry, and complexity of the information (Sener et al. 2019). Thus, approaches to data mining should be comprehensive, staged, and consistent models that organize information and provide the basis for certain conclusions. There are three major approaches: Cross Industry Standard Process for Data Mining (CRISP-DM), “Sample, Explore, Modify, Model and Assess” (SEMMA) and Knowledge Discovery in Databases (KDD). According to Schmitt and Schuh (2018), these three models are more focused on extracting illustrative information than on practical application to modifying production and management processes. Thus, they primarily perform a descriptive function, which may be the basis for further actions.
KDD is the process of finding useful knowledge in raw data. This knowledge may include rules describing the relationships between data properties (decision trees), frequently encountered patterns (association rules), as well as the results of classification and data clustering. Schmitt and Schuh note (2018) that KDD “is too specialised on assessing databanks,” which makes it less applicable to the actual processes in the company’s SCM (p. 54). This approach is the most theoretical and abstract and is suitable for a wide range of data.
Contrary to KDD, CRISP-DM and SEMMA are based on a cyclical approach and are intended for regular analysis of the company’s logistic processes. SEMMA, as the name implies, includes five steps and focuses on the modelling objectives without covering the business aspects. This approach is positioned as a unified cross-sectoral approach to the iterative process of intelligent data analysis (Schmitt & Schuh 2018). According to Kharlamov, Ferreira and Godsell (2019), “comparatively CRISP-DM has been suggested as being more company-oriented than the very popular KDD, and it is also more complete than SEMMA” (p. 5). It consists of six stages, including business understanding, data understanding, data preparation, modelling, evaluation and deployment. Every stage is divided into intermediate points, which “represent milestones to determine the suitability of the chosen process for the project goal” (Schmitt & Schuh 2018, p. 54). This approach is initially focused on the commercial goals of the company and allows considering data mining projects not as theoretical computations, but as a complete element of business processes.
Descriptive Data Mining Techniques
These approaches describe the general outline of the analytical process, but do not have a strict methodology and allow a variety of data mining techniques. They perform a descriptive function, and as a rule, their use implies applying corresponding methods of data analysis. Researchers note that “for descriptive analytics, the association is the most widespread as it has been applied throughout every stage of the supply chain process” (Nguyen et al. 2018, p. 260). This technique detects associations between data items and reveals underlying patterns within the set. Data mining techniques such as classification, clustering and regression can also be used at certain stages of each approach. They are intended for categorising data based on specific criteria, transforming data for further analysis, finding data groups similar in specific attributes, and detecting correlations between different variables.
Analysed Areas of Logistics and SCM
These approaches may be applied to a wide range of SCM goals. According to Nguyen et al. (2018), the literature examines the Big data analysis application to both the optimisation of routine processes and the analysis of data in real-time and prospective decision making. It should be noted that the above approaches are most suitable for optimising regular processes as they handle arrays of stable data and reveal significant patterns. Wang et al. (2016) claim that data mining approaches are applicable to both SCM strategies (strategic sourcing, supply chain network design and product design development”) and operations (“demand planning, procurement, production, inventory and logistics”) (p. 102). For instance, CRISP-DM can be used for consistent analysis of the overall supply chain design data within and outside the company. At the same time, this approach can be applied to the analysis of inventory point operations. Thus, the scope of data mining approaches is only limited by the discretion of the analysts due to their flexible methodology.
Conclusion
It should be noted that the major descriptive data mining approaches are KDD, SEMMA and CRISP-DM. Among all approaches, CRISP-DM is the most appropriate for analysing the SCM data because it is more business-oriented than the KDD and more comprehensive and structured than the SEMMA. Various descriptive data mining techniques may be employed within these approaches, such as association, classification, clustering and regression. The given approaches and techniques are applied in a wide spectrum of SCM areas for the descriptive analysis of regular logistic processes.
Reference List
Kharlamov, A. A., Ferreira, L. M. D. F. and Godsell, J. (2019) ‘Developing a framework to support strategic supply chain segmentation decisions: a case study’, Production Planning & Control, 30, pp. 1-14.
Schmitt, R. and Schuh, G. (eds.) Advances in production research: Proceedings of the 8th Congress of the German Academic Association for Production Technology (WGP), Aachen, 2018. Cham: Springer.
Nguyen, T. et al. (2018) ‘Big data analytics in supply chain management: A state-of-the-art literature review’, Computers & Operations Research, 98, pp. 254-264.
Sener, A. et al. (2019) ‘The role of information usage in a retail supply chain: A causal data mining and analytical modeling approach’, Journal of Business Research, 99, pp. 87-104.
Wang, G. et al. (2016) ‘Big data analytics in logistics and supply chain management: Certain investigations for research and applications’, International Journal of Production Economics, 176, pp. 98-110.