Research Projects

OM & Financial Engineering

·         Natural Gas Trading Strategies with Capacitated Storage Facilities and High Dimensional Futures Price Model, with Nicola Secomandi, François Margot.

This project studies the optimal (physical) trading strategy with a capacitated storage facility and a high-dimensional futures price model. We calibrate the futures price model from real historical data (NYMEX) building the price model under forward measure and construct the close-optimal trading policy based on an approximate dynamic program. This policy is superior to those heuristics used in practice for natural gas trading. Moreover, the dynamic program also produces a very tight upper bound which can provide guidance for traders to measure the suboptimality of the trading strategies as well as to value the storage facility. The trading system (with numerical algorithms and simulation system) is developed under C++/Linux.

·         Examination of Futures Price Models for Natural Gas Commodities with Physical Trading and Delta Hedging, with Nicola Secomandi, Duane Seppi, Alan Scheller-Wolf, François Margot.

This project studies the futures price models for the natural gas commodity. In particular, we compare the effectiveness of the number of factors included in a model from the viewpoints of both accuracy and efficiency. Existing research has evaluated price models by financial instruments (for instance, using delta hedging and call/put options to compare the discrepancy due to the price model). However, there exist various issues related to the physical trading purpose, for instance, physically buying-in and selling-out natural gas with storage limits. In this project, we utilize the optimization model for physical trading developed in the above project and examine the effectiveness of the price model by both physical trading and financial trading (delta hedging). We evaluate the discrepancy due to the price model and propose evaluation matrix by both historical data and simulation. The system is developed under C++/Linux.

·         Bond Financing, Capacity Investment and Production with Correlated Demand Stream and Market Interest Rate

We study the decisions on capacity investment and bond financing combined with the sequential production strategy for a firm with limited capital. We model the market demand by an autoregressive time series correlated with the market interest rate. The interest rate process is implemented by a discrete CIR model. We study the production strategy under given facility capacity level and capital structure (cash and debt level). We provide algorithm to simulate the expected value of the firm and determine the default risk. We determine the at-market bond price. This framework considers a firm’s detailed operations to examine the bond default risk, which differs from Black-Scholes structural (or Moody’s KMV EDF) credit risk model. Numerical algorithm and simulation system are developed under C++.

OM & Corporate Finance

·         Stock Market Pressure, Sales Reporting, Inventory Signal and Inventory Investment for Publicly Traded Firms, with Laurens Debo, Lin Nan

We consider operational decisions on inventory investment and sales reporting of publicly traded firms that bear stock market pressure. Management and ownership are separated in those firms. We study inventory management decisions with correlated demand. The manager’s compensation is partially based on the firm’s stock price which is influenced by the reported inventory level and the reported sales. With better information about the demand, the manager may manipulate the stock price by investing more or less inventory, shipping more inventory than the real demand to downstream customers and reporting higher than real sales revenues. We reveal the incentives that drive the manager’s sales manipulation and address potential differences of public firms’ operational decisions from private firms. [see “Channel Stuffing” on Wiki; email me for real business cases]

·         Financing, Inventory Risk and Supply Contracting, with Laurens Debo, Katia Sycara

We study inventory investment for firms with financial constraint within a supply chain. The firms’ internal capital may not be sufficient to support desired operations and thus may need to finance from a capital market. The firms bear bankruptcy risk which determines the cost of financing. We study the firms’ optimal capital structure and the optimal supply chain design with respect to the inventory risk allocation.

OM & Marketing

·         Strategic Consumer Behavior, Price Matching Policy, Pricing and Inventory Investment, with Laurens Debo, Katia Sycara, Cuihong Li

Sellers may price high at the beginning of the sale and mark down later with excess inventory. However, strategic consumers expect the markdown possibilities and may wait to buy later, which impacts a seller’s profit and then decreases the inventory investment. With a posterior price-matching policy, a seller guarantees to refund the price difference in the case of markdown. We examine the impact of such a policy on consumers’ purchasing behavior. We show that this policy is able to eliminate waiting incentive. We find that this policy can improve the seller’s inventory investment as well as profit substantially if the fraction of strategic consumers is not too small and their valuation decline is neither too low nor too high. Moreover, it can also increase consumer surplus when the market variance is high.

Supply Chain Management & Practice

·         Supply Chain Management Platform for Sunpu Electronics Co.(Beijing, China), with Yueting Chai, Yi Liu and other group members (in Tsinghua University)

In this project, we study, design and develop a supply chain management (software) system for Sunpu Electronics Co. Ltd., Beijing, China. The system is used to manage corporate-level supply chain information, including procurement, order processing, production planning, transportation, consumer relationship management, supply relationship management, etc. I focused on workflow management and inventory decision. I developed the system with colleagues (based on Visual studio and Oracle database).

Multiagent Systems & Automated Negotiation

Automated negotiation allows an autonomous software agent to negotiate with other agents or humans on sales, contracting, task allocation, etc., which is an important mechanism for Ecommerce and Multi-Agent applications. In this project, we explore a generic framework with different negotiation schemes for automated multi-attribute negotiations. The framework incorporates the issues of: generality of the utility functions, incomplete information and Pareto-optimality. The framework also considers tractability. We implemented our work for the U.S. Navy detailing and training process.

In the censor network project, we explore a general framework and algorithms for coordinating large scale teamwork applied for Air force search munitions. We organize team members into dynamically evolving sub-teams and design token-based information sharing scheme. We develop domain independent software proxies with which we demonstrate teams at least an order of magnitude bigger than previously studies. In another ongoing work, we explore how to organize the censor network in a distributed manner with stochastically arriving tasks.