Abstract:
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.
