We are back with Dr. Diadie Sow for the third and final part of our series dedicated to Data Science. The first part was dedicated to the story of Data Science up to now, and the second part focused on the daily tasks of a Data Scientist. In this article, we will explore the value that Data Science brings to Supply Chain. Spoiler alert: huge innovations ahead!
What is the role of data science in the supply chain?
In today’s supply chain management systems, machine learning algorithms implemented by data scientists are used at every step of the strategic planning process. This includes production planning, procurement planning, and goes all the way up to demand planning. At QAD Dynasys, we have been integrating these machine learning solutions step by step in our demand planning module to help customers improve their demand forecast accuracy, as well as with their data cleansing and customer segmentation. We have carried out several such projects with our customers, particularly in the food and beverage and the automotive industries. These projects are sometimes carried out on datasets drawn directly from the QAD DynaSys’ DSCP solution, and other times using a customer’s specific dataset. An example of this could be Point of Sales data sales and production order documents from distribution centers and production plants. The projects’ results were very positive and helped us to automate several important new functionalities of DSCP that are centered on data cleansing, customer segmentation, and demand forecasting.
How does Data science revolutionize the supply chain industry?
The supervised, unsupervised, or reinforced machine-learning algorithms used by data scientists have brought about a major revolution in all parts of supply chain processing.
“One of the main purposes of supply chain collaboration is to improve the forecast accuracy”(Raghunathan, 1999)
In the demand planning role, Machine learning has proven satisfactory results in improving accuracy of forecasting new products or products with short history, but also products with complex seasonality etc. Indeed, Machine learning forecast is not only based on the historical data like traditional statistical forecasting methods, but it also takes into account causal factors and this may influence demand yet had been previously unknown, or looking for the behaviors of other products with similarities, already present on the market.
On the other hand, Machine learning visual pattern recognition is also proving to be very effective on the supply chain process management. For example, “The machine learning algorithms in IBM’s Watson platform were able to determine if a shipping container and/or product were damaged, classify it by damage time, and recommend the best corrective action to repair the assets. Watson combines visual and systems-based data to track, report and make recommendations in real-time.”(Artificial Intelligence in Logistics, 2018 (PDF, 45 pp., no opt-in)
What advice would you give a student wanting to get into this line of work?
According to “The Times”, data science is one of the “attractive” fields of work in the twenty-first century, with a wealth of new opportunities and huge job creation. Consequently, there has been a large influx of people wanting to work in data science. Some are so enthusiastic about the field that they are inspired to leave their current jobs and retrain in data science.
The best advice I could offer is to keep it up in order to make their dreams come true. This requires a very strong background in both math and in programming. A doctorate in mathematics is not compulsory, even though it may be preferred or even a prerequisite for the bigger American companies and in Anglo-Saxon countries. However, a perfect understanding of differential calculus, optimization, probability, statistics, and linear algebra is necessary. Today, most machine learning algorithms are available in user-friendly packages, but without the knowledge of math, it is difficult to understand how the algorithm works; this can lead to real issues when faced with real-life problems. To understand the reasons behind the poor performance of a specific dataset or to choose the most appropriate algorithm, it is absolutely necessary to understand the math behind it all.