2051. Rethinking the Container Allocation Problem from a Machine Learning Perspective
Invited abstract in session MC-4: Data science meets strongly NP-Hard CO, stream Data Science meets Optimization.
Monday, 12:30-14:00Room: Rupert Beckett LT
Authors (first author is the speaker)
| 1. | Kavitha Chetana Didugu
|
| Data and Analytics, Zooplus SE |
Abstract
While consulting for one of the largest Shipping Yard Companies in India, I took a different perspective to solve the Container Allocation Problem. Problem: Currently, the incoming containers were being haphazardly arranged. This causes high retrieval cost when customers come back to collect their containers. They were looking for an easy to implement, and easy to scale (across multiple yard locations across India) method for yard allocation that would reduce the retrieval cost. Retrieval cost is directly proportional to the number of picks or moves required by the crane (called reach-stacker) to retrieve a stacked container at the time of retrieval request.
Traditionally, this problem has been solved using an Optimisation (Operations Research) approach. Given the amount of complexity in building and executing this solution, I explore a novel ML approach to solve this problem. This breaks down a complex optimisation problem into a forecasting model. The ML approach offers the advantage of higher accuracy with larger data, faster execution time, and easy scalability.
This application-centred paper delves deeper into how to reformulate the container allocation problem within the ML framework, and further steps in the solution. For a set of unseen incoming container data, I compared the results of the company's existing plan versus the plan obtained from my method, and observed that my plan reduced the cost of operations by 66.67%. Additionally, the model also achieved a
Keywords
- Machine Learning
- Artificial Intelligence
- Logistics
Status: accepted
Back to the list of papers