Flexsim for Data Analytics & Machine learning

[asb_praneeth]

Flexsim has good connectivity with a lot of databases. It has great dashboards and power querying interfaces(calculated tables). Why can't Flexsim develop a module for data analytics and machine learning?

[benjamin_w2]

Hi @asb.praneeth,

Have you looked into FlexSim's optimizer? It is the closest thing to machine learning that FlexSim has. In general, FlexSim is designed to generate data rather than analyze it. However, the optimizer uses ML techniques to find the optimal configuration for a given target. Check out this link.

https://docs.flexsim.com/en/19.1/Reference/Tools/ExperimenterOptimizer/

If you have any ideas for specific ML applications for FlexSim, let us know.

[asb_praneeth]

Yes. I know that. But I want to say that Why can't Flexsim have some data analytics and machine learning capabilities as standalone. This will expand the Flexsim applications to wide range of applications.

Even Simulation is a complex function of inputs. Instead of running a simulation for hours, Is there a way we just train on the inputs and with simulation output and approximate a function that predicts the Output for a new set of inputs. - Just a Idea. May not be 100% correct.

[braydn_t]

I'm not sure I understand what you are getting at. Why would you want approximate solutions for a model (which would likely take more time to train than to run the optimizer because you would have to train it on your specific model) When you could get a more correct solution by just running your model? There are also very few models that take hours to run. You can set up flexsim to export the results of your model run, and turn the speed up all the way and simulate months of time in minutes.

Again, I don't think I really understand what the advantage would be over the Optimizer/Expirementer.

[sam_stubbsYXX86]

For information about this, look in the user manual.  Go to: Building the Model's Logic > Using Reinforcement Learning.