Modern decision analytics frequently involves the optimization of an objective over a finite horizon where the functional form of the objective is unknown. The decision analyst observes covariates and tries to learn and optimize the objective by experimenting with the decision variables. We present a nonparametric learning and optimization policy with covariates. The policy is based on adaptively splitting the covariate space into smaller bins (hyper-rectangles) and learning the optimal decision in each bin.

We synthesize the field of machine learning with the canonical problem of empirical asset pricing: measuring asset risk premia. In the familiar empirical setting of cross section and time series stock return prediction, we perform a comparative analysis of methods in the machine learning repertoire, including generalized linear models, dimension reduction, boosted regression trees, random forests, and neural networks. At the broadest level, we find that machine learning offers an improved description of expected return behavior relative to traditional forecasting methods.

We study a service system which does not have the capability of monitoring and disclosing its real-time congestion level. However, the customers can observe and post their observations online, and future arrivals can take into account such user-generated information when deciding whether to go to the facility. We perform pairwise comparisons of the shared, full, and no queue length information structures in terms of social welfare.

In this paper, we offer a new network perspective on one of the central topics in Operations Management - demand variability and the bullwhip effect (BWE). The topic has both practical and scholarly implications. We start with a puzzling observation: while the traditional intra-firm measure of the BWE increases at upstream layers of the network, the demand variability experienced by upstream firms actually decreases. To reconcile these two facts, we hypothesize that firms manage their customer base to smooth out the aggregate demand.

Expanding on the zero-inflated Poisson (ZIP) model, the multiple-inflated Poisson (MIP) model is applied to analyze count data with multiple inflated values. The existing studies on the MIP model determined the inflated values by inspecting the histogram of count response and fitting the model with different combinations of inflated values, which leads to relatively complicated computations and may overlook some real inflated points.

We propose a distribution-free approach to solving a moral hazard model in which a max-min principal hires an agent who selects the outcome distribution subject to moment constraints. Our formulation reveals that the model has an alternative interpretation of two-sided ambiguity where the principal and agent have opposing robust decision rules but form congruent expectations on the distribution selected by nature.

This paper aims to reduce the length of stay (LOS) in Emergency Departments (EDs) by designing a systematic response policy for various specialists depending on the demands of their consultation. We model the specialist consultation (SC) demands via non-homogeneous Poisson process of a daily cycle; and then based on the martingale representation theorem, we figure out the optimal SC start times for a Fixed Time (FT) policy, in order to minimize the average per-person wait time.

In many services, a customer's valuation for the service depends on the amount of time that customer requires for service. We thus consider a queueing model to explicitly capture such dependence in congestion-prone services. Our goal is to study the impact of the dependence on the service provider's pricing decision and revenue performance. We first consider a benchmark case in which customers are charged a fixed fee for service.

The paper examines statistical and economic evidence of out-of-sample bond return predictability for a real-time Bayesian investor who learns about parameters, hidden states, and predictive models over time. We find some statistical evidence using information contained in forward rates. However, such statistical predictability cannot generate any economic value for investors. Furthermore, strong statistical and economic evidence from fully revised macroeconomic data vanishes when real-time and survey-based macroeconomic information is used.

Over the last couple of years, Bitcoin cryptocurrency and the Blockchain technology that forms the basis of Bitcoin have witnessed an unprecedented attention. Designed to facilitate a secure distributed platform without central regulation, Blockchain is heralded as a novel paradigm that will be as powerful as Big Data, Cloud Computing, and Machine Learning. Blockchain continues to evolve, but its applications have already matured to rival, and already in some cases, replace more traditional institutions as avenues of global activity.