Second-level Decision Research Articles

Two-step model based on XGBoost for predicting artwork prices in auction markets

Art markets globally have grown, making artwork an investment of note. Precise valuation is pivotal for optimal returns. We introduce a two-step model with a two-level regressor, utilizing extreme gradient boosting (XGBoost) for accurate artwork price prediction. The model encompasses a price-class classifier and regressors for individual categories. This captures diverse factor influences, combining predictions to reduce misclassification risks. Visual features further enhance accuracy through the second-step two-level regressor. Experiments on Korean art auction data demonstrate the superiority of our two-step model with the two-level regressor over one-step and two-step alternatives, as well as the hedonic pricing model. While visual features affected one- and two-step models’ training, they boosted performance when integrated into the second-level decision tree, reducing first-level residuals. This emphasizes the two-level regressor’s efficacy in incorporating visual elements for artwork valuation. Our study highlights the potential of our approach in the field of artwork valuation.

Exact algorithms for solving a bi-level location\u2013allocation problem considering customer preferences

The issue discussed in this paper is a bi-level problem in which two rivals compete in attracting customers and maximizing their profits which means that competitors competing for market share must compete in the centers that are going to be located in the near future. In this paper, a nonlinear model presented in the literature considering customer preferences is linearized. Customer behavior means that the customer patronizes the most attractive (most comfort) location that he/she wants to be served among the locations of the first-level decision maker (Leader) and the second-level decision maker (Follower). Four types of exact algorithms have been introduced in this paper which include three types of full enumeration procedures and a developed branch-and-bound procedure. Moreover, a clustering-based algorithm has been presented that can provide a good approximation (a good lower bound) to the mentioned binary problem. For this purpose, the numerical results obtained are compared with the results of the full enumeration, heuristic and the branch-and-bound procedure.

Piecewise affine parameterized value-function based bilevel non-cooperative games

Generalizing certain network interdiction games communicated to us by Andrew Liu and his collaborators, this paper studies a bilevel, non-cooperative game wherein the objective function of each player’s optimization problem contains a value function of a second-level linear program parameterized by the first-level variables in a non-convex manner. In the applied network interdiction games, this parameterization is through a piecewise linear function that upper bounds the second-level decision variable. In order to give a unified treatment to the overall two-level game where the second-level problems may be minimization or maximization, we formulate it as a one-level game of a particular kind. Namely, each player’s objective function is the sum of a first-level objective function ± a value function of a second-level maximization problem whose objective function involves a difference-of-convex (dc), specifically piecewise affine, parameterization by the first-level variables. This non-convex parameterization is a major difference from the family of games with min-max objectives discussed in Facchinei et al. (Comput Optim Appl 59(1):85–112, 2014) wherein the convexity of the overall games is preserved. In contrast, the piecewise affine (dc) parameterization of the second-level objective functions to be maximized renders the players’ combined first-level objective functions non-convex and non-differentiable. We investigate the existence of a first-order stationary solution of such a game, which we call a quasi-Nash equilibrium, and study the computation of such a solution in the linear-quadratic case by Lemke’s method using a linear complementarity formulation.

Ad-Blockers: A Blessing or a Curse?

Users who have an ad-blocker installed present a genuine predicament for a website (a.k.a. publisher): On the one hand, these users do not generate revenue for the website; on the other hand, denying them access can shrink the user base and adversely affect the popularity of the website, ultimately reducing traffic over the long run. This has led some websites to require that ad-block users “white-list” them for obtaining access to an “ad-light” experience. We model the decision problem for a website facing two user segments: regular users and ad-block users. The first-level decision or gating strategy is whether to allow ad-free access to ad-block users or require them to white-list the website for gaining access. When ad-block users are allowed ad-free access, the second-level decision is the level of advertising (or ad intensity) for regular users. When ad-block users are required to white-list, the second-level decisions are the ad-intensities for regular users and ad-block users. The net utility of a user from visiting the website depends on the intrinsic value of the website’s content, the value obtained due to network effects driven by the amount of traffic/popularity of the website, and the cost incurred due to the presence of ads. We derive an optimal gating and ad-intensity strategy for the website and also solve an identical model for a world without ad-block software. We show that the website can increase its revenue by discriminating between regular and ad-block users via the ad-intensities shown to them. More interestingly, we find that the discriminatory power bestowed on the website by ad-blockers can also benefit its users when their outside option is not very attractive. Thus, the advent of ad-blockers can lead to a win-win for both the website and its users. Finally, we propose a superior selective-gating strategy in which only a fraction of ad-block users are gated. We establish the robustness of our conclusions under several enhancements to our base setting: (a) heterogeneous profitabilities from regular users and ad-block users, (b) endogenous adoption of ad-blockers, (c) the presence of a subscription option, and (d) negative externality due to increased traffic. Our analysis ends with recommendations for three stakeholders in this problem, namely, publishers, web-browser developers, and policy makers.

MULTI-LOOK SAR ATR USING TWO-LEVEL DECISION FUSION OF NEURAL NETWORK AND SPARSE REPRESENTATION

As for the lack of the contribution by decision fusion in pose estimation and the demand for the combination of the feature fusion and the decision fusion in SAR ATR, in this paper, with the help of pose estimation, a new multi-look SAR ATR method is proposed in order to improve the performance, which is based on two-level decision fusion of neural network and sparse representation. The first-level decision fusion is acted for the combination of the pose estimation result by neural network and sparse representation. Based on the constraint of pose, these two models are exerted for the multi-look SAR ATR, and the second-level decision fusion is used to achieve the final recognition result. Several experiments based on MSTAR are conducted, and experimental results show that our method can achieve an acceptable result.

A PSO-based procedure for a bi-level multi-objective TOC-based job-shop scheduling problem

This study presents an application of particle swarm optimisation (PSO) algorithm for a bi-level multi-objective job-shop scheduling problem. The bi-level decision-making requirement stems from the concept of theory of constraints. At the first level, the decision is made by concentrating on minimising idle time on the system bottleneck. The second-level decision is made to plan other machines while maintaining the maximum use of the bottleneck and gaining improvements in other performance measures. This paper proposed a PSO-based procedure for solving the bi-level programming problem. The proposed procedure simplifies the solution method by simultaneously providing solutions for the objective of both levels. In addition, during the schedule generation process, the job-shop case applied in this study also considers the machine set-up time, transfer lot size and product demands to make the model more realistic. The numerical examples are given to demonstrate how this approach works. The results from this procedure are compared with the solutions obtained by a commercial optimiser, the LINGO 10 software package. This proposed PSO is implemented in C# programming language in order to obtain the final solution within the short computational time.

Minimizing Labor Requirements in a Periodic Vehicle Loading Problem

In this paper, we address a logistics problem that a manufacturer of auto parts in the north of Spain described to the authors. The manufacturer stores products in its warehouse until customers retrieve them. The customers and the manufacturer agree upon an order pickup frequency. The problem is to find the best pickup schedule, which consists of the days and times during the day that each customer is expected to retrieve his/her order. For a given planning horizon, the optimization problem is to minimize the labor requirements to load the vehicles that the customers use to pick up their orders. Heuristically, we approach this situation as a decision problem in two levels. At the first level, customers are assigned to a calendar, consisting of a set of days with the required frequency during the planning horizon. Then, for each day, the decision at the second level is to assign each customer to a time slot. The busiest time slot determines the labor requirement for a given day. Therefore, once customers have been assigned to particular days in the planning horizon, the second-level decision is a multiprocessor scheduling problem, where each time slot is the equivalent of a processor, and where the objective is to minimize the makespan. A metaheuristic procedure is developed for the problem of minimizing labor requirements in this periodic vehicle-loading problem and artificial as well as real data are used to assess its performance.

Hierarchical Decision Making when the Second-level Decision Maker Works According to Reward and the Related Incentive Strategy

This paper deals with the hierarchical decision making process when the second-level decision maker works according to reward. We first propose a Stackelberg game model with ringlike structure to describe such decision making processes, use the "sub-team optimal solution" to indicate the best outcome to the leader, and then give several methods to design incentive strategies which can induce "sub-team optimal solution", and a numerical example is given to illustrate the results of the paper at the end.

An overview of the first and second level trigger of DELPHI

The DELPHI trigger is organized in four levels. The first two triggers are synchronous with the beam crossover, every 22 mu s, and are implemented in hardware. The first-level trigger decision is taken at 3 mu s and is used primarily as a fast trigger to gate the time protection chamber. Gating this detector off for rejected events reduces space-charge buildup. If the first-level decision is positive, the following beam crossover will be lost to allow the inclusion of data from long-drift-time detectors in the second-level trigger. The second-level decision is taken at 39 mu s. If this decision is positive a front-end readout is initiated and up to 500 mu s is required to free the front-end buffers. To limit the deadtime fraction for each trigger to the order of 2% the first-level trigger rate has to be held to 1 kHz and the second-level rate to 20 Hz. The authors describe the various subtriggers in DELPHI together with the detector local trigger decision boxes contributing to the subtrigger, and they indicate the type of hardware used implement the decision logic. >

Leader-follower strategies for multilevel systems

Sequential strategies for dynamic systems with multiple decision makers and multiple performance indices are surveyed and reviewed. These strategies are generalizations of Stackelberg or leader-follower strategies for two-person games. The review includes structures with one coordinator and several second-level decision makers, and linear hierarchical structures with only one decision maker at each level Several information structures are considered.

Computer Identification of Constrained Handprinted Characters with a High Recognition Rate

Handprinted character recognition by computer is accomplished on forty-nine character classes with a high recognition rate (> 99.4 percent). The form of characters is constrained by requiring each character to be handprinted on a standard grid. The grid is composed of twenty line segments, each of which forms the basis for a feature, yielding twenty features to represent each character. A person printing these characters is not expected to remain precisely on the grid lines. The errors that do occur in following the grid lines are assumed to be normally distributed; therefore, each is based on a longitudinal Gaussian-shaped surface. A page of constrained characters to be recognized is input to the computer using a television camera. Each character on the page is located, isolated from the other characters, and quantized into binary points. A novel technique utilizing a Gaussian point-to-line distance concept for calculation of feature value has been employed. The recognition program extracts the twenty values and attempts to determine in which of the forty-nine character classes the unknown character belongs. This decision is made based on these twenty features using a weighted minimum distance classifier. If only a marginal classification can be made, a second-level decision is used to increase the likelihood of correct classification. The second-level decision uses the most discriminating features of the two most likely character classes in order to increase the likelihood of correct classification. All character-dependent data are obtained through training techniques.