Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 105 - Deep Learning and Statistical Modeling with Applications
Type: Invited
Date/Time: Monday, August 3, 2020 : 1:00 PM to 2:50 PM
Sponsor: International Chinese Statistical Association
Abstract #308073
Title: Statistical Challenges in Analyzing Two-Sided Marketplace
Author(s): HONGTU ZHU*
Companies: DiDi
Keywords: causal inference; deep learning; reinforcement learning; two sided market; MDP; deep Q-network
Abstract:

In this talk, we introduce a general AI for two sided market place analytical framework. This framework integrates some fundamental statistical methods, including experiment design and causal inference, with modern machine learning methods, such as deep learning and reinforcement learning. This framework not only makes statistical inference on a particular policy, but also optimizes a set of policies for a complex system. For instance?we introduce a novel class of equilibrium metrics (EMs) to quantify spatial balance of dynamic demand and supply networks defined on the same graph. It is primarily motivated by measuring the local and global spatial coherence between demand and supply patterns in large- scale ride-sharing platforms, such as Uber. Moreover, we establish the causal inference framework for dynamic process in order to introduce a set of statistical methods for evaluating various polycies used in ride sharing business. In addition, we model the ride dispatching problem as a Markov Decision Process and propose learning solutions based on deep Q-networks with action search to optimize the dispatching policy for drivers on ride-sharing platforms.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program