Multi-armed bandit analysis for price optimization

Question

Recently, I read a blog post "Bandits Know the Best Price of a Product", ( http://pkghosh.wordpress.com/2013/08/25/bandits-know-the-best-product-price/ ), which describes how to use multi-armed bandit analysis to optimize prices.

There is also a lot of discussion about whether multi-armed bandit analysis is better than A / B testing (for example, “20 lines of code that will beat A / B testing every time”): http://stevehanov.ca/blog /index.php?id=132?utm_medium=referral vs. "Why is a multi-armed bandit algorithm not" better "than testing A / B": http://visualwebsiteoptimizer.com/split-testing-blog/multi-armed-bandit -algorithm / ).

I know that there is an R-package called a bandit that can be used for such an analysis.

Does anyone have an example of a game - comparable to the one on the blog - that shows how to apply this method using R ( in the context of price optimization )?

Thank you for your help.

Answer 1

Recently, I have been doing projects with gangster algorithms. Basically, the performance of bandit algorithms is determined by a lot of data. And it is very good for continuous testing with data change. So what you need to do to test and configure your model when testing data.

To take more bandits, you can read this book, bandit algorithms for website optimization: http://shop.oreilly.com/product/0636920027393.do . He explains the basic gangster algorithms well enough and implements them in Python. You can find his code on Github: https://github.com/johnmyleswhite/BanditsBook . However, they did not talk about contextual bandits in the book.

For R, I'm not sure. But I just searched the Internet, I found a guy who implemented gangsters in R, here is the code: https://github.com/lotze/bandit

Hope this helps you.

Answer 2

My careful research on this topic may come in handy: http://codeandmath.wordpress.com/2014/04/05/type-i-error-in-bandits/