Machine Learning as an improvement mechanism for optimization problems

User 3166 | 2/14/2016, 5:06:54 AM

I have been thinking about ways, how we could use machine learning as an improvement mechanism for allocation.

Let's say we have a company that operates a fleet of buses on different routes. Now as a start they could want a cost function based optimiser that can spit out a bus to route assignment. That would be a case, if this were a bus company which had just setup shop, but in the case of a bus company that has been operating for a while, they would have a ton of historical data already about the buses that have operated routes in the past when they did not have an automated system. In this case it would be possible to use a machine learner to actually function as an optimiser learning from this manual operational assignment data.

I would appreciate if someone could shed some light on this use case. How could one go about using machine learners to act as optimisers in task allocation scenarios like, bus to route allocation, resource to project task allocation, load balancing of server load etc.

Comments

User 1592 | 2/14/2016, 4:05:48 PM

Hi Shajee, There are many possible approaches depends on the exact problem. For example, for demand forecasting of bikes sharing transportation system you can view this example: https://dato.com/learn/gallery/notebooks/kagglebikeshare_prediction.html

Optimizing routes is a different problem and there graph analytics may be helpful.

I will be happy to setup a time to discuss further the problem details.