Ramakrishnan Pakath

A unified foundation for business analytics

Synthesizing prior research, this paper designs a relatively comprehensive and holistic characterization of business analytics - one that serves as a foundation on which researchers, practitioners, and educators can base their studies of business analytics. As such, it serves as an initial ontology for business analytics as a field of study. The foundation has three main parts dealing with the whence and whither of business analytics: identification of dimensions along which business analytics possibilities can be examined, derivation of a six-class taxonomy that covers business analytics perspectives in the literature, and design of an inclusive framework for the field of business analytics. In addition to unifying the literature, a major contribution of the designed framework is that it can stimulate thinking about the nature, roles, and future of business analytics initiatives. We show how this is done by deducing a host of unresolved issues for consideration by researchers, practitioners, and educators. We find that business analytics involves issues quite aside from data management, number crunching, technology use, systematic reasoning, and so forth. Holistic foundation of business analytics introduced, as an ontology for the fieldCharacterization business analytics as evidence-based problem recognition and solvingAdvance a framework of business analytics for a cohesive unifying view of the fieldStimulates thinking about nature, role and future of business analytics initiativesDeduces unresolved issues for considerations by researchers/educators/practitioners

A genetics-based hybrid scheduler for generating static schedules in flexible manufacturing contexts

Existing computerized systems that support scheduling decisions for flexible manufacturing systems (FMSs) rely largely on knowledge acquired through rote learning for schedule generation. In a few instances, the systems also possess some ability to learn using deduction or supervised induction. We introduce a novel AI-based system for generating static schedules that makes heavy use of an unsupervised learning module in acquiring significant portions of the requisite problem processing knowledge. This scheduler pursues a hybrid schedule generation strategy wherein it effectively combines knowledge acquired via genetics-based unsupervised induction with rote-learned knowledge in generating high-quality schedules in an efficient manner. Through a series of experiments conducted on a randomly generated problem of practical complexity, we show that the hybrid scheduler strategy is viable, promising, and, worthy of more in-depth investigations. >

Decision making under time pressure with different information sources and performance-based financial incentives: part 1

We are witness to the communications revolution and the accompanying proliferation of narrow-purpose, mobile, computing and communication devices. Such devices tend to be smaller and lighter than their desktop and laptop counterparts. The tradeoff is that their displays and memory also tend to be relatively smaller. To date, they also rely on traditional English and/or icons for communicating with users. While icons have grown in usage, capturing any and all information using icons is impossible and/or prohibitively expensive. We examine the viability of developing new kinds of communication languages for such devices in a specific setting by considering an abstract classification task and examining the performance of subjects using a new, compact language that we have devised vis-a-vis written and spoken English. Our work draws on prior research on induced value experimentation and ex-ante system evaluation. In Part 1 of this two-part paper, we provide the necessary background, discuss the underlying motivations, and describe the construction and refinement of our experimental platform and an accompanying subject training software suite.

Learning by problem processors: adaptive decision support systems

In this paper, we describe the potential advantages of developing Adaptive Decision Support Systems (Adaptive DSSs) for the efficient and/or effective solution of problems in complex domains. The problem processing components of DSSs that subscribe to existing DSS paradigms typically utilize supervised learning strategies to acquire problem processing knowledge (PPK). On the other hand, the problem processor of an Adaptive DSS utilizes unsupervised inductive learning, perhaps in addition to other forms of learning, to acquire some of the necessary PPK. Thus, Adaptive DSSs are, to some extent, self-teaching systems with comparatively less reliance on external agents for PPK acquisition. To illustrate these notions, we examine an application in the domain concerned with the scheduling of jobs in flexible manufacturing systems (FMSs). We provide an architectural description for an Adaptive DSS for supporting static scheduling decisions in FMSs. We illustrate key problem processing features of the system using an example. A prototype system, based on this architecture, is currently under implementation.

Four models for a decision support system

Abstract We examine four decision support system (DSS) models – the Symbiotic, Expert, Holistic, and Adaptive – and distinguish them in terms of the impact of their knowledge management styles on their problem-processing behavior. We draw upon existing notions of knowledge types and their management to develop a knowledge-oriented view. We use it to categorize the models as being either Static or Dynamic. From this perspective, the Holistic DSS may be regarded as being the most advanced, as it postulates holistic problem recognition and processing capabilities. While progress has been made on digitally simulating holistic recognition, much remains to be done in developing practical processors and truly holistic systems that couple such processors and recognizers.

The Iterated Prisoner's Dilemma: early experiences with Learning Classifier System-based simple agents

Abstract Prior research on artificial agents/agencies involves entities using specifically tailored operational strategies (e.g., for information retrieval, purchase negotiation). In some situations, however, an agent must interact with others whose strategies are initially unknown and whose interests may counter its own. In such circumstances, pre-defining effective counter-strategies could become difficult or impractical. One solution, which may be viable in certain contexts, is to create agents that self-evolve increasingly effective strategies from rudimentary beginnings, during actual deployment. Using the Iterated Prisoners Dilemma (IPD) problem as a generic agent-interaction setting, we use the Learning Classifier System (LCS) paradigm to construct autonomously adapting “simple” agents. A simple agent attempts to cope by maintaining an evolving but potentially perennially incomplete and imperfect knowledge base. These agents operate against specifically tailored (non-adaptive) agents. We present a preliminary suite of simulation experiments and results. The promise evidenced leads us to articulate several additional areas of interesting investigations that we are pursuing.

Searching for information in a time-pressured setting: experiences with a Text-based and an Image-based decision support system

Searching for the right information and making quick, accurate decisions within time-pressured settings is often non-trivial. We contrast the relative efficacies of written English (Text) and a more concise, compact communication mode (Image) for information search and decision making by using a financial incentive scheme to apply implicit time pressure on subjects. We found that, while Image users earned as much as Text users, they achieved this earnings parity by following speedier but less accurate strategies. We conclude with thoughts on possible refinements to our work that could steer subjects in the ideal direction of fast, accurate, lucrative decisions with languages like Image.

Optimal buffer inventories for multistage production systems with failures

Recently, several procedures have been developed to solve the problem of optimal location and sizing for buffer inventories in a serial production system that is subject to random variations in the magnitude or timing of stage failures, repair times, and demands for finished goods. To handle production systems of realistic complexity, we have developed a simplified mathematical model and an efficient solution procedure based on dynamic programming. The solution procedure has been implemented in a portable, public-domain computer program which can also be applied to nonserial systems with diverging branches. We illustrate the procedure using a hypothetical diverging-branch system with 15 stages.

A Formal Approach for Designing Distributed Expert Problem-Solving Systems

In this paper, we consider the problem of generating effective information-gathering, communication, and decision-making ICD strategies for a distributed expert problem-solving DEPS system. We focus on the special case of a dual-processor DEPS system and present a decision-theoretic model that enables the characterization of feasible, efficient, and optimal ICD strategies. In view of the tremendous amount of computing needed to generate optimal strategies for problems of practical size, we develop useful heuristic procedures for constructing high-quality efficient ICD strategies. We illustrate the use of the model and the solution procedure through an example.

An examination of evolved behavior in two reinforcement learning systems

Using agent-based simulation experiments, we assess the relative performance of two Reinforcement Learning System (RLS) paradigms - the classical Learning Classifier System (LCS) and an enhancement, the Extended Classifier System (XCS) - in the context of playing the Iterated Prisoners Dilemma (IPD) game. In prior research, the XCS outperforms the LCS in solving the Animats-and-Maze and Boolean Multiplexer test problems. Our work has overlaps with and is an extension of such efforts in that it allows assessment of each systems ability to (a) cope with delayed environmental feedback, (b) evolve irrational choice as the optimal behavior, and (c) cope with unpredictable input from the environment. We find that while the XCS is considerably superior to the LCS, in terms of four key performance metrics, in playing IPD games against a deterministic, reactive game-playing agent (Tit-for-Tat), the LCS does better against an unpredictable opponent (Rand) albeit with significant evolutionary effort. Further, upon examining each XCS enhancement in isolation, we see that specific LCS variants equipped with a single XCS feature, do better than the traditional LCS model and/or the XCS model in terms of particular metrics against both types of opponents but, again, usually with greater evolutionary effort. This suggests that if offline, rather than online, performance and specific performance goals are the focus, then one may construct relatively-simpler LCS variants rather than full-fledged XCS systems. Further assessments using LCS variants equipped with combinations of XCS features should help better comprehend the synergistic impacts of these features on performance in the IPD.