Designed to meet business needs for efficient scheduling, Yumbo is an advanced solution with a robust mathematical model that optimizes task scheduling with precision and flexibility. Let's take a closer look at its key features.
The foundation of Yumbo’s mathematical model was laid with operations research techniques, ensuring that business problems were formulated with mathematical rigor.
The model was built from the ground up using linear programming, where the objective function was designed to favor early task completion, as required by managers.
To meet real-world scheduling needs, integer linear programming constraints were introduced to shape the scheduling logic and business requirements.
The program introduced a unique approach: designing schedules by focusing on constraints. This allowed for greater flexibility and adaptability to changing business needs.
Model was implemented using AMPL, integrating AMPL Python modules to enhance computational capabilities.
The design of the model focused on test cases with 1,000 experts, tasks, and constraints that required memory-efficient solutions. Two different models were introduced: one with and one without UBDAY constraints to provide different memory consumption profiles.
Support for HiGHS, GCG, and SCIP solvers was added, streamlining efficient optimizations. An interesting finding was that different solvers often produced different schedules with identical objective function values.
A major enhancement was the introduction of overflow tasks model, which allowed the system to generate schedules even when the number of assigned experts was insufficient for timely task completion.
💡 As Yumbo continues to evolve, organizations can leverage its powerful mathematical optimization techniques to efficiently solve complex scheduling challenges.
🔗 Yumbo is available as an open-source tool on GitHub https://github.com/romz-pl/yumbo/ and as ready-to-run system https://yumbo-ampl.streamlit.app/ .
#OperationsResearch
#Optimization
#DataScience
#Streamlit
#Python
#AMPL
#LinearProgramming
Image created by ChatGPT.
