In this post, we will explore some applications of clustering algorithms within Logistics Management space that can actually be applied relatively easily and are actually already being used by many organizations.
For those who want to try building these models in Python, please email me on firstname.lastname@example.org and I can share the Python code and sample input dataset, that uses K Means clustering for Inventory classification . The sample data is made up data but you can use the code on scaled up data as well.
1. Inventory classification :
Companies that have to deal with thousands of SKUs always have a challenge categorizing/classifying new SKUs into any classification that they follow (like A,B,C classification). Clustering algorithms can be leveraged to categorize incoming SKUs. grouping inventories by value, manufacturing and sales margins etc.
This type of solution can then be married with an automated Inventory optimization solution. So everytime a new SKU is added, the algorithm automatically assigns it a class and the IO tool then leverages that class to apply an inventory policy (class based) to perform safety stock calculations.
2. Warehouse design and picking optimization:
This analytical approach is based around the concept of grouping highly correlated inventory together and placing them in the warehouse as close as possible. Two algorithms that are tried and tested in this scenario are Principal Component Analysis and Singular Value Decomposition . After inventory clustering, you can then use an assignment algorithm to assign locations to parts in the facility in a way that minimizes the total travel distance.
There is another approach to improving picking efficiency, that combines cluster and association analysis to improve order picking efficiency.
This is also a two step process in which first a clustering algorithm is used to categorize the types of goods in orders. Classification can be performed based on multiple criterias, like the turnover of goods, value, sales volume, favorable commodity ratings, whether free shipping is provided and whether cash on delivery is supported etc..
Next, an association algorithm is used to determine the relationships among goods by studying the habits of consumers who buy them. A method for improving the class-based storage strategy is proposed. The picking distance of the improved storage strategy can be compared with that of the traditional strategy via simulation experiments.
3. Route Optimization:
Since Route optimization problems are heuristics based, they are generally difficult to solve and hence time and (computing) resource intensive.
Studies have shown that leveraging a clustering algorithms to cluster demand nodes as an initial step reduces the total solve time significantly. In a route optimization run experiment that I recently did, I ran a route optimization problem in an off the shelf tool and then ran the same small sample data using an open source VRP stack. In the VRP stack, the heuristics first uses K-means clustering algorithm to get local distribution centers and customer points within their scope, and then ant colony algorithm to find the optimal distribution route. In multiple runs, the VRP Stack run time was always 30%+ less than the off the shelf tool.