Learning project on public data.
How much does a linear program save over a rule of thumb?
I solved the transportation and transshipment models from MIT 15.057, then applied the same method to the ChemCo network from Houston and Dallas.
Transportation LPAgainst northwest corner15.3%121,450 against 143,400
Transshipment LPAgainst a greedy plan2.1%4,210 against 4,300 ($000)
ChemCo LPVariable freight a yearUS$999KAs is: US$1.01M
Close DallasAll direct from HoustonUS$1.87MUS$861K more freight
What I built
- Solved each model with PuLP and checked the shadow prices by solving again.
- Built a network view of flows and a sensitivity chart for the Houston to Dallas rate.
- Explained every shadow price in plain words.
- Added formula cards for the LP, the transshipment balance and the cost per gallon.
What the numbers show
The LP beats the rule of thumb in both teaching models. On ChemCo, routing each customer the cheaper way saves US$12K a year against the current setup, and Dallas keeps most gallons until the round trip rate passes US$3,690.