Terry R. Rakes

Using Autonomous Software Agents to Create the Next Generation of Decision Support Systems

The purpose of this research is to explore the promise of autonomous software agents in Decision Support Systems (DSS). Because definitions of software agents extant in the literature are divergent, we develop and provide a descriptive definition useful for our purpose. The benefits of agents and the particular characteristics of agents leading to DSS enrichment are examined. To facilitate this we build a DSS described elsewhere in the literature and enhance it with different types of autonomous software agents. From this experience, a general framework for agent-enabled DSS is suggested. It is concluded that such a DSS in general will be more difficult to build than traditional DSS, but at least some agent-enabled DSS will bring significant benefit to users.

The effect of sample size and variability of data on the comparative performance of artificial neural networks and regression

This research explores the robustness of simple linear regression and artificial neural networks with respect to varying sample size and variance of the error term by comparing their predictive abilities. The comparison is made using the root mean square difference between the predicted output from each technique and the actual output.

Going the last mile: a spatial decision support system for wireless broadband communications

High-speed, wireless communication networks are increasing in popularity; however they can be costly and difficult to plan. In this paper we present a spatial decision support system that incorporates expert knowledge of wireless communications, area topography, demographics and propensity to pay for service in order to aid wireless network planners determine optimal placement of equipment to maximize profit or minimize cost. Moreover, the system can be useful in performing policy analysis to determine pricing, governmental subsidy levels, etc. By integrating a GIS tool into the DSS, planners can easily adjust parameters to better understand the problem at hand and move toward bringing broadband connectivity to the last mile.

A mathematical model for scheduling and staffing multiclinic health regions

Abstract This paper describes the design and analysis of a multiobjective integer linear program for scheduling and staffing multiple clinics with itinerant health personnel in multiclinic regions. Specifically, the optimal assignment of physicians, nurse practitioners, nurses and/or nursing assistants to differing types of medical clinics with differing personnel needs is considered in conjunction with determination of individual clinic operating schedules. The scheduling model incorporates parameters such as distance between clinics, personnel availability, personnel time, demand for services, and the mix of personnel required to meet quality goals, minimize travel costs, and maximize staff preferences for certain assignments. The model is illustrated using data provided by family planning decision makers in a rural health care environment. Analysis of the results shows that the formulation is easily generalizable to multiple settings and that the essence of the scheduling decision can be represented by the mathematical programming framework. Because of the size of the problem, techniques for improving computational aspects are discussed. The model is designed to be used as a benchmark to enable the development of heuristic scheduling methods which can be used on a PC for near optimal scheduling.

Single machine scheduling: A comparison of two solution procedures

Recent research on the single machine scheduling problem has focused on the treatment of multiple scheduling objectives. Most works have used some combination of mean flowtime, maximum tardiness, or total tardiness as scheduling criteria. Previous research has largely ignored earliness as a scheduling criterion. This paper presents a model that employs the criteria of flowtime as a measure of work-in-process (WIP) inventory and total job earliness to represent finished goods inventory. Total tardiness is used to represent customer satisfaction. The three criteria are used to form a single, weighted-sum objective function for guiding the choice of the best processing sequence. Two procedures are presented that might be used to solve this problem. The first is an enumeration scheme using bounding and dominance criteria that have been developed to aid efficient solution, and the second is a mixed integer linear programming (LP) formulation. Computational experience with the two models is also presented.

Neural network identification of critical factors in a dynamic just-in-time kanban environment

Prior research has examined the proper number of kanbans to be used in various just-in-time environments, but relatively little work has been done in exploring which factors internal and external to a shop in a given time period are critical in determining the necessary number of kanbans to be specified for the next period. The research reported here examines the identification of shop factors in a dynamic and stochastic just-in-time environment. In particular, three questions are addressed: does information from a prior period help in setting the kanban level in the current period? If so, which endogenous and exogenous factors considered individually help the most? And finally, what grouping of individual factors is most important in deciding the number of kanbans? The methodology employed is to use artificial neural networks to fit simulated shop data to learn the relationship between prediction factors and overall shop performance. Appropriate non-parametric statistical tests are then used to answer the questions. The answers obtained, although shop specific, may also be generated by firms willing to follow the procedure presented here for conditions specific to their particular operation.

Automated merging of conflicting knowledge bases, using a consistent, majority-rule approach with knowledge-form maintenance

This paper discusses an automated process of merging conflicting information from disparate sources into a combined knowledge base. The algorithm provided generates a mathematically consistent, majority-rule merging by assigning weights to the various sources. The sources may be either conflicting portions of a single knowledge base or multiple knowledge bases. Particular attention is paid to maintaining the original rule format of the knowledge, while ensuring logical equivalence. This preservation of rule format keeps the knowledge in a more intuitive implication form as opposed to a collection of clauses with many possible logical roots. It also facilitates tracking using the support for each deductive result so that final knowledge in rule form can be ascribed back to original experts. As the approach is fairly involved mathematically, an automated procedure is developed.

Using Autonomous Software Agents in Decision Support Systems

The purpose of this research is to define autonomous software agents, and describe a general framework for the use of agents in decision support systems (DSS). Because definitions of software agents extant in the literature are divergent, we develop and provide a descriptive definition useful for our purpose. The benefits of agents and the particular characteristics of agents leading to DSS enrichment are examined. To facilitate this we build a DSS, described elsewhere in the literature, and enhance it with different types of autonomous software agents. From this experience, a general framework for agent-enabled DSS is suggested.

An equity approach to contractor assignment in post-disaster debris disposal operations

Disaster debris cleanup typically accounts for nearly one-third of the total cost of post-disaster recovery. In the USA, state and local municipalities rely on federal financial assistance administered by the US Federal Emergency Management Agency (FEMA). Ineffective decision-making and allocation of resources can severely strain financial resources and cause social and political unrest. In addition to obvious objectives such as cost minimisation and FEMA suggestions related to contractor logistics, Disaster Management Coordinators (DMCs) should consider equitable cleanup across regions as an important objective when deciding how to allocate resources. This paper presents a multiple objective mixed-integer decision model for assisting decision makers in allocating resources in support of disaster debris cleanup operations. The model incorporates the unique assumptions, objectives, and constraints of post-disaster debris cleanup. The effectiveness of the proposed model is demonstrated using data from debris cleanup operations in Chesapeake, Virginia, following Hurricane Isabel in 2003.

Using Autonomous Planning Agents to Provide Model-based Decision-making Support

The purpose of this research is to investigate the potential of software agents in performing automated planning within Decision-Making Support Systems (DMSS). A planning agent that uses a partial-order planning mechanism to monitor and maintain corporate profit is built in Prolog and Java and embedded in an agent-integrated DMSS. The planning mechanism used by the agent supports the sequencing and interfacing of models within the DMSS for the purpose of generating alternatives. From this experience, an architecture for a general planning agent is constructed. Benefits of the agent-enhanced, planning DMSS are then discussed.