Machine learning is not a new topic nor is it new to supply chain. However, it has been garnering a lot of attention within supply chain due to its transformational business potential. Whatis.com defines machine learning as “a type of artificial intelligence (AI) that allows software applications to become more accurate in predicting outcomes without being explicitly programmed. The basic premise of machine learning is to build algorithms that can receive input data and use statistical analysis to predict an output value within an acceptable range.”
As defined by expertsytems.com machine learning is often categorized as being supervised and unsupervised. Supervised machine learning algorithms can apply what has been learned in the past to new data using labeled examples to predict future events. Starting from the analysis of a known training dataset, the learning algorithm produces an inferred function to make predictions about the output values. The system is able to provide targets for any new input after sufficient training. The learning algorithm can also compare its output with the correct, intended output and find errors in order to modify the model accordingly. Machine learning algorithms are considered unsupervised when the information used to train them is neither classified nor labeled. Unsupervised learning studies how systems can infer a function to describe a hidden structure from unlabeled data. The system doesn’t figure out the right output, but it explores the data and can draw inferences from datasets to describe hidden structures from unlabeled data.
The applications for machine learning within supply chains are numerous. Machine learning is helping demand planners to generate more accurate forecasts by analyzing historical sales and correlating external events, and consumer behavior, to peaks and troughs in demand. Traditional forecasting solutions focus on identifying patterns within the sales data itself and require trained statisticians to tune the engine for improved performance. Machine learning algorithms, on the other hand, focus more on the independent variables (i.e., weather, nearby events, competitor price actions, etc.) and attempt to find correlations between these external events and the dependent variable — sales. By finding correlated independent variables which may be leading indicators to future demand, machine learning algorithms can produce more accurate forecasts than traditional solutions. As an added bonus, machine learning algorithms can be unsupervised, continuously exploring and discovering new correlations, and hence improving accuracy.
Another high value application for machine learning in supply chain relates to lead time and throughput variability. Think of the demand forecasting example, but instead of trying to predict sales, machine learning algorithms attempt to predict the performance of each supplier, carrier, forwarder, port, lane, road, manufacturing facility, warehouse, etc. within the extended supply chain, under varying conditions. Have certain carriers on specific lanes shown a propensity to make unscheduled port stops (resulting in delays) when spot rates are more attractive, for example. What do dwell times at the port of Rotterdam look like when 15 vessels show up at the same time vs. 30? How long does it take for a warehouse to unload a truck when its dock doors are 75% full and it has fourteen dock staff working? By constantly analyzing these performance metrics under varying conditions, machine learning algorithms begin to establish learned behavior models that can lead to very accurate predicted times of arrival (PTAs) as well as variability. With accurate PTAs and lead time / throughput variability predictions, organizations can significantly reduce inventory levels because they no longer have to offset uncertainty.
Supply chain practitioners often complain that it’s not a lack of data that is hampering them; it’s too much data. Machine learning is helping companies to overcome this problem by converting data overload into value creation.