Tony Wu

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2016-57

May 12, 2016

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2016/EECS-2016-57.pdf

We present this thesis to study and build automated pricing engines that work in on-demand economies. The task of determining market prices is especially challenging in an on-demand marketplace because of frequently fluctuating supply and demand. An automated pricing agent provides reactive real-time prices. To evaluate different types of automated pricing models, we build an on-demand marketplace simulation. By using research on human behavior, we create customer and supplier agent models that interact within the simulation. Afterwards, we run market experiments using various pricing algorithms. We introduce the concept of applying reinforcement learning to generate dynamic prices, and we evaluate our reinforcement learning agent's performance relative to that of other simpler pricing algorithms. Performance is measured not only in terms of profit, but also in metrics that determine customer and supplier retention.

Advisors: Anthony D. Joseph

---

BibTeX citation:

@mastersthesis{Wu:EECS-2016-57,
    Author= {Wu, Tony},
    Editor= {Joseph, Anthony D. and Russell, Stuart J.},
    Title= {Automated Pricing Agents in the On-Demand Economy},
    School= {EECS Department, University of California, Berkeley},
    Year= {2016},
    Month= {May},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2016/EECS-2016-57.html},
    Number= {UCB/EECS-2016-57},
    Abstract= {We present this thesis to study and build automated pricing engines that work in on-demand economies. The task of determining market prices is especially challenging in an on-demand marketplace because of frequently fluctuating supply and demand. An automated pricing agent provides reactive real-time prices. To evaluate different types of automated pricing models, we build an on-demand marketplace simulation. By using research on human behavior, we create customer and supplier agent models that interact within the simulation. Afterwards, we run market experiments using various pricing algorithms. We introduce the concept of applying reinforcement learning to generate dynamic prices, and we evaluate our reinforcement learning agent's performance relative to that of other simpler pricing algorithms. Performance is measured not only in terms of profit, but also in metrics that determine customer and supplier retention.},
}

EndNote citation:

%0 Thesis
%A Wu, Tony 
%E Joseph, Anthony D. 
%E Russell, Stuart J. 
%T Automated Pricing Agents in the On-Demand Economy
%I EECS Department, University of California, Berkeley
%D 2016
%8 May 12
%@ UCB/EECS-2016-57
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2016/EECS-2016-57.html
%F Wu:EECS-2016-57