Replies to This Discussion

Permalink Reply by Lance Olson on August 18, 2012 at 2:26pm

As much as I like game theory, I think that machine learning and predictive analytics would bring more of a benefit to everyone.  Game theory, in general, presupposes that people are going to act rational and only think about a few alternatives.   Some industries there are too many players and too many variables to implement game theory effectively, like retail, wholesale, and grocery.   Too many consumer preferences to creep into the mix, there are a lot of alternatives and preferences in those areas which in turn make game theory very complex in depicting the interaction. 

For the industries that have relatively low number of competitors and/or low number of alternatives then I think game theory could be used effectively.  Some industries that come to mind: telecommunications, automobile, housing (buyer/seller), public utilities and some others.

• ▶ Reply

Permalink Reply by rajendra on August 18, 2012 at 2:49pm

Machine learning and predictive analysis are the tools. No qualms whatsoever. But avant-garde machine-learning is at the mercy of "optimization" and "scheduling". We need a 'greedy solution' to start with and only hypothesis that we have at our disposal. 

"strategic step towards a local goal is impeded by the local constraints cum global constraints"- a game theory perspective

In reality we can think about few alternatives at a time. All analytic techniques are assuming people to act rationally. 

• ▶ Reply

Permalink Reply by Lance Olson on August 18, 2012 at 3:29pm

Rajendra,

I agree with you for the most part...not that you need my approval.  :)

The only part that I am having a hard time with is your last statement, maybe you can shed some light on the perspective. 

"All analytic techniques are assuming people to act rationally. "  - Think large numbers for the sample.  We do not care who shows up to buy a product just so long as someone does.   Or we do not know specifically who will get in an accident but someone will.   Maybe that is not the perspective that I should be reading into your sentence.

I do think that game theory is basically a subset of predictive analytics tools, without understanding the interactions among the agents we will have a hard time performing predictive analytics.  The interactions and the details of those interactions may or may not play a role into the solution as a whole in predictive analytics.  It does give us greater understanding in general of what is "necessary" and/or "unnecessary."  

Great post.

• ▶ Reply