Four families of machine learning, each with typical algorithms, a plain definition and a supply chain example. Use it to match the method to the decision you need to make.
Swipe the table sideways to read every column.
| AI/ML Method | Algorithms | Generic Definition | Application Example in Supply Chain and Logistics |
|---|---|---|---|
| Supervised Learning | Linear Regression, Logistic Regression, Support Vector Machines (SVM), Random Forest, Gradient Boosting, Neural Networks | A method where the algorithm is trained on labeled data, and the goal is to learn a mapping function from inputs to outputs. | Predicting the delivery time of a shipment based on historical data of delivery times and other relevant factors such as distance, weather, and traffic. |
| Semi-Supervised Learning | Label Propagation, Expectation-Maximization (EM) Algorithm, Co-Training | A method where the algorithm is trained on a combination of labeled and unlabeled data, and the goal is to improve the accuracy of the model using the unlabeled data. | Detecting anomalies in inventory levels by analyzing historical data of inventory levels and identifying unusual patterns. |
| Unsupervised Learning | K-Means, Hierarchical Clustering, Principal Component Analysis (PCA), Autoencoders | A method where the algorithm is trained on unlabeled data, and the goal is to discover hidden patterns and structures in the data. | Clustering similar products together in a warehouse based on their attributes such as size, weight, and fragility. |
| Reinforcement Learning | Q-Learning, SARSA, Deep Reinforcement Learning | A method where the algorithm learns by trial and error through interactions with the environment, and the goal is to maximize a reward signal. | Optimizing the routing of delivery trucks to minimize the total distance traveled while ensuring timely deliveries. |
A guided tour of the main machine-learning families used in logistics: regression, classification, clustering and reinforcement learning.
Each method is shown with a typical supply chain use case, its data needs, and its failure modes.
Match the method to the decision: forecast with regression, segment with clustering, sequence decisions with RL.
