Novel Methods for Adjudication of Multiple Cognitive Decision Models
نویسندگان
چکیده
Decision Support Systems (DSS) derive information from ingested data streams and present it to cognitive agents (usually human users) for exploitation. DSS generally have multiple data sources available, each having its own characteristics. Users must select from among these data sources based upon their perceived processing needs. This we call the Information Source Selection (ISS) Problem. It is a fundamental component of the broader fusion problem. In systems where the availability, relevance, complexity, etc., of information sources is dynamic, users must reselect information sources when conditions change. This is problematic, because ISS is a complex, situation-dependent, combinatorial problem. Done poorly, it misuses valuable information. Done well, it makes best-use of data by identifying for the user which sources are likely to be helpful and unhelpful; this enables the user to maintain system performance while controlling cognitive complexity in space and time. The ISS process described here uses results tabulated during an automated sampling procedure to estimate the joint contribution of sets of information sources in context. This automation allows a decision support system to perform first-cut ISS for a user automatically, suggesting a set of information sources optimized for current or emerging situations. These cooperating sets of jointly-effective information sources we call “source cliques”. We describe the underlying algorithm, and report the results of empirical experiments showing that an ISS method based upon a Markov Random Field and Gibbs Sampling [1] can automatically and effectively select a source set for a given cognitive processing goal. Empirical results are given showing that cliques corresponding to Gibbs energy minima give superior classifier performance, while cliques of unselected features substantially under-perform.
منابع مشابه
A novel model of clinical reasoning: Cognitive zipper model
Introduction: Clinical reasoning is a vital aspect of physiciancompetence. It has been the subject of academic research fordecades, and various models of clinical reasoning have beenproposed. The aim of the present study was to develop a theoreticalmodel of clinical reasoning.Methods: To conduct our study, we applied the process of theorysynthesis in accordan...
متن کاملTarget setting in the process of merging and restructuring of decision-making units using multiple objective linear programming
This paper presents a novel approach to achieving the goals of data envelopment analysis in the process of reconstruction and integration of decision-making units by using multiple objective linear programming. In this regard, first, we review inverse data envelopment analysis models for data reconstruction and integration. We present a model with multi-objective linear programming structure in...
متن کاملMultiple attribute decision making with triangular intuitionistic fuzzy numbers based on zero-sum game approach
For many decision problems with uncertainty, triangular intuitionistic fuzzy number is a useful tool in expressing ill-known quantities. This paper develops a novel decision method based on zero-sum game for multiple attribute decision making problems where the attribute values take the form of triangular intuitionistic fuzzy numbers and the attribute weights are unknown. First, a new value ind...
متن کاملA hybrid decision making system using DEA and fuzzy models for supplier selection in the presence of multiple decision makers
متن کامل
Efficiency evaluation of wheat farming: a network data envelopment analysis approach
Traditional data envelopment analysis (DEA) models deal with measurement of relative efficiency of decision making units (DMUs) in which multiple-inputs consumed to produce multiple-outputs. One of the drawbacks of these models is neglecting internal processes of each system, which may have intermediate products and/or independent inputs and/or outputs. In this paper some methods which are usab...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2006