An application on supply chain management with big data and machine learning methods

Supply chain management is a discipline that covers the entire process from the supply of raw materials that make up a product to the returns from consumers. Today, with the effect of digitalization and infrastructure studies, supply chains have left the classical structure and transformed into a structure that contains high amounts of data and technological equipment is used. In this context, sensors have been used intensively at every stage in the supply chain network. With the intense use of sensors, classical supply chain managements have begun to be named as internet of things (IoT) based supply chain management. In IoT-based supply chain management, sensors are used at every stage and equipment, from the vehicles used in transportation to the shelves where the products are sold. With these sensors, instant data is obtained and big data about the whole process emerges. Machine learning, on the other hand, is artificial intelligence-based algorithms developed to analyze the large amount of data obtained. When the supply chain management literature is examined, it is seen that there is a lack of studies that deal with the chain as a whole and include machine learning and analysis. The results of this study contain important findings when the topic is up-to-date and the gap in the literature is considered. In this context, real data on the raw material procurement process and production process were obtained by considering the supply business of a manufacturing company. This big data obtained was analyzed with machine learning algorithms on Microsoft Azure Machine Learning Studio platform. In the first part, supply delays were determined by applying linear regression and forecasts for future periods were developed. In the second part, firstly, regression analysis was applied to determine the significant relationships between the data on the supply chain process and to determine the correlation between the variables to be the basis for the next stages, and to determine the effect on the dependent variable. In the continuation, a feed forward artificial neural network model was developed and the actual production levels were compared with the production levels predicted by machine learning, and the average absolute errors were determined and the production efficiency of the enterprise was expressed. In the last part, a hypothetical linear programming model was developed and the problem of the enterprise's production and delivery of the products it stored to five different distribution centers was solved with the LINDO package program. It is estimated that the results obtained and the algorithms used will set an example for the sector and fill the gap in the field. Finally, suggestions for future research are presented.