Optimizing Inventory Routing: An Integrated Machine Learning Solution Approach
This is the master thesis study. It provides a novel solution approach for an inventory routing problem (IRP).
The solution approach is mat-heuristic that utilizes both a mathematical model and machine learning algorithm.
Multi-layer perceptrons are used in the machine learning part which helps to significantly reduce time spent for sub-procedures.
An earlier paper has also been benchmarked and improvements are presented.
Experimental parts of the study has been completed. It is aimed to be submitted in 2024
Key contributions of the paper can be summarized as:
- A promising machine learning pipeline to be used in a combinatorial optimization problem
- A mathematical model which is also adoptable to many similar IRP settings
- A better performing algorithm than benchmarks
- Feature engineering specific to IRP-type problems