Soft Computing Evaluation Logic: The Lsp Decision Method and Its Applications

ISBN-13: 9781119256458 / Angielski / Twarda / 2018 / 912 str.

Jozo Dujmovic

Opis

This book provides a methodology for decision making problems developed in the context of soft computing, fuzzy systems and computational intelligence. The book includes four parts that cover an introduction to evaluation, the evaluation logic theory, the Logic Scoring of Preference (LSP) method, and applications. Part One (Evaluation Decision Problems) is an introduction to evaluation. Part Two (Evaluation Logic) presents a detailed theoretical background for the weighted idempotent aggregation of suitability and modeling of mental evaluation processes. Part Three (LSP Method) presents the area of professional evaluation, the design of complex evaluation logic criteria, and all other components of the LSP method. Part Four (Applications) includes a spectrum of applications of the LSP method in the areas of system evaluation, comparison, selection, and optimization.

Spis treści

Preface xviiAbout the Companion Website xxiiiPrevious Publications xxivAcknowledgments xxvList of Symbols and Abbreviations xxviiPart One EVALUATION DECISION PROBLEMS 11.1 Intuitive Evaluation as a Logic Decision Process 51.1.1 Main Observable Steps of the Intuitive Evaluation Process 61.1.2 Subjective and Objective Components in Evaluation 181.2 Quantitative Evaluation--An Introductory Example 211.2.1 Stakeholders and Their Goals 211.2.2 Attributes 221.2.3 Attribute Criteria 231.2.4 Simple Direct Ranking 271.2.5 Aggregation of Attribute Suitability Degrees 291.2.6 Using Cost and Suitability to Compute the Overall Value 321.3 Drawbacks of Simple Additive and Multiplicative Scoring and Utility Models 351.3.1 Simple Additive Scoring: The Irresistible Attractiveness of Simplicity 361.3.2 Simple Multiplicative Scoring 451.3.3 Logic Unsuitability of Scoring and Utility Theory Models in Professional Evaluation 471.4 Introduction to Professional Quantitative Evaluation 511.4.1 Five Fundamental Types of Professional Evaluation Problems 511.4.2 A Survey of Typical Professional Evaluation Problems 541.4.3 Components of Methodology for Professional Quantitative Evaluation 58Part Two GRADED LOGIC AND AGGREGATION 632.1 Graded Logic as a Generalization of Classical Boolean Logic 692.1.1 Aggregators and Their Classification 702.1.1.1 Means 712.1.1.2 General Aggregation Functions 712.1.1.3 Logic Aggregators 732.1.1.4 Triangular Norms and Conorms 732.1.2 How Do Human Beings Aggregate Subjective Categories? 752.1.3 Definition and Classification of Logic Aggregators 852.1.4 Logic Bisection, Trisection, and Quadrisection of the Unit Hypercube 922.1.5 Propositions, Value Statements, Graded Logic, and Fuzzy Logic 952.1.6 Classical Bivalent Boolean Logic 1002.1.7 Six Generalizations of Bivalent Boolean Logic 1082.1.7.1 Expansion of Function Domain 1092.1.7.2 Expansion of Logic Domain 1112.1.7.3 Expansion of Annihilator Adjustability 1122.1.7.4 Expansion of Semantic Domain 1152.1.7.5 Expansion of Compensative Logic Functions 1172.1.7.6 Expansion of the Range of Andness/Orness from Drastic Conjunction to Drastic Disjunction 1182.1.8 GL Conjecture: Ten Necessary and Sufficient GL Functions 1232.1.9 Basic Idempotent GL Aggregators 1272.1.10 A Summary of Differences between Graded Logic and Bivalent Boolean Logic 1342.1.11 Relationships between Graded Logic, Perceptual Computing, and Fuzzy Logic 1362.1.12 A Brief History of Graded Logic 1422.2 Observable Properties of Human Evaluation Logic 1472.2.1 Perceptual Computer and Its Basic Properties 1522.2.2 Simultaneity and Substitutability in Evaluation Models 1772.2.3 Basic Semantic Aspects of Evaluation Logic Reasoning 1902.2.4 Multipolarity: Grouping and Aggregation of Semantically Heterogeneous Inputs 2122.2.5 Grouping and Aggregation of Semantically Homogeneous Inputs 2182.2.6 Imprecision, Incompleteness, Logic Inconsistency, and Errors 2222.3 Andness and Orness 2372.3.1 A General Definition of Andness/Orness 2372.3.2 Local Andness and Orness in the Simplest Case of Two Variables 2392.3.3 Variability of Local Andness 2422.3.4 Mean Local Andness and Orness in the Case of Two Variables 2482.3.5 Local and Mean Local Andness and Orness in the Case of n Variables 2512.3.6 Global Andness and Orness 2532.3.7 Mean Global Andness/Orness Theorems and Their Applications 2722.3.8 Geometric Interpretations of Andness and Orness 2752.4 Graded Conjunction/Disjunction and Logic Modeling of Simultaneity and Substitutability 2832.4.1 Definitions and Basic Mathematical Properties of Logic Aggregators 2842.4.2 Classification of Conjunctive and Disjunctive Logic Aggregators 2952.4.3 Properties of Means Used in Logic Aggregation 2982.4.4 Algebraic Properties of Aggregators Based on Weighted Power Means 3042.4.5 Logic Aggregators Based on Weighted Means with Adjustable Andness/Orness 3132.4.6 Selection and Use of the Threshold Andness Aggregator 3182.4.7 Andness-Directed Interpolative GCD Aggregators 3272.4.8 Uniform and Nonuniform Interpolative GCD Aggregators 3342.4.8.1 The Uniform Interpolative GCD Aggregator (UGCD) 3342.4.8.2 An Extremely Soft Interpolative Aggregator 3382.4.8.3 An Extremely Hard Interpolative Aggregator 3382.4.9 Extending GCD to Include Hyperconjunction and Hyperdisjunction 3422.4.10 From Drastic Conjunction to Drastic Disjunction: A General GCD Aggregator 3472.4.11 Gamma Aggregators versus Extended GCD Aggregators 3482.4.11.1 Multiplicative and Additive Gamma Aggregators 3512.4.11.2 Comparison of Gamma Aggregators and GCD 3552.4.12 Four Main Families of GCD Aggregators and Sixteen Conditions They Must Satisfy 3612.5 The Percept of Importance and the Use of Weights 3672.5.1 Multiplicative, Implicative, and Exponential Weights as Importance Quantifiers 3692.5.1.1 Multiplicative Weights 3702.5.1.2 Implicative Weights and the Weighted Conjunction/Disjunction 3742.5.1.3 Exponential Weights 3902.5.2 Impact of Weights on Aggregation Results 3932.5.3 Semantic Components in Logic Aggregation Models 3982.5.4 Seven Techniques for Weight Adjustment 4022.5.4.1 Importance Decomposition Method 4022.5.4.2 Direct Weight Assessment 4052.5.4.3 Weights Based on Ranking 4052.5.4.4 Weights Based on Menu 4072.5.4.5 Collective Weight Determination 4092.5.4.6 Weights Obtained from Pairwise Comparisons 4112.5.4.7 Weights Based on Preferential Neuron Training 4142.5.5 Multivariate Weighted Aggregation Based on Binary Aggregation Trees 4172.6 Partial Absorption: A Fundamental Asymmetric Aggregator 4292.6.1 Conjunctive Partial Absorption 4302.6.2 Disjunctive Partial Absorption 4362.6.3 Visualizing the Partial Absorption Function, Penalty, and Reward 4392.6.4 Mathematical Models of Penalty and Reward 4422.6.5 Selecting Parameters of Partial Absorption 4492.7 Logic Functions That Use Negation 4532.7.1 Negation and De Morgan's Duality 4532.7.2 De Morgan's Laws for Weighted Aggregators and Dualized Weighted Aggregators 4552.7.3 De Morgan's Duals of Compound Functions 4582.7.4 Nonidempotent Logic Functions 4602.8 Penalty-Controlled Missingness-Tolerant Aggregation 4632.8.1 Missing Data in Evaluation Problems 4632.8.2 Penalty-Controlled Numerical Coding of Missing Data 4652.8.3 A Penalty-Controlled Missingness-Tolerant Aggregation Algorithm 4672.8.4 The Impact of Penalty on Missingness-Tolerant Aggregation 4722.9 Rating Scales and Verbalization 4752.9.1 Design of Rating Scales 4762.9.1.1 Strict Monotonicity of Linguistic Labeling 4772.9.1.2 Linearity of Rating Scales 4832.9.1.3 Balance of Rating Scales 4862.9.1.4 Cardinality of Rating Scales 4882.9.1.5 Hybrid Rating Scales 4892.9.2 Stepwise Refinement of Rating Scales for Andness and Orness 4912.9.3 Scaling and Verbalizing Degrees of Importance 4962.9.4 Scaling and Verbalizing Degrees of Suitability/Preference 497Part Three LSP METHOD 4993.1 An Overview of the LSP Method 5013.1.1 Characterization of Stakeholder and Organization of an Evaluation Project 5033.1.2 Development of the Suitability Attribute Tree 5063.1.3 Elementary Attribute Criteria 5143.1.4 Logic Aggregation of Suitability 5193.1.4.1 Logic Aggregation Using Graded Conjunction/ Disjunction 5233.1.4.2 Logic Aggregation Using Partial Absorption 5263.1.5 Cost/Suitability Analysis and Comparison of Evaluated Objects Using Their Overall Value 5363.1.6 Summary of Properties of the LSP Method 5403.2 LSP Decision Engineering Framework for Professional Evaluation Projects 5433.2.1 Participants in a Professional Evaluation Process Based on LSP DEF 5443.2.2 Relationships between Evaluators and Domain Experts 5463.2.3 The Structure of LSP DEF and the Corresponding Professional Evaluation Process 5473.2.4 Predictive Nature of Evaluation Models 5513.2.5 Interpretation of Evaluation Results 5523.2.6 Complexity, Completeness, and Accuracy of Evaluation Models 5533.2.7 Combining Opinions of n Experts 5553.2.7.1 The Maximum Likelihood Estimate 5553.2.7.2 The Expert Competence Estimate 5573.3 Elementary Attribute Criteria 5613.3.1 Notation of Elementary Criteria 5613.3.2 Verbalization of Elementary Criteria 5653.3.3 Continuous Nonlinear Elementary Criteria 5663.3.4 Classification of Twelve Characteristic Types of Elementary Criteria 5693.4 Aggregation Techniques and Tools 5793.4.1 Selecting GCD Aggregators for an LSP Project 5793.4.2 Selecting GCD Aggregators by Training Preferential Neurons 5813.4.3 Analytic Techniques for Selecting Partial Absorption Aggregators 5893.4.3.1 AH Version of the Conjunctive Partial Absorption Aggregator 5893.4.3.2 AH Version of the Disjunctive Partial Absorption Aggregator 5943.4.4 Boundary Penalty/Reward Tables for Selecting Partial Absorption Aggregators 5953.4.5 Selecting Partial Absorption Aggregators by Training Preferential Neurons 5973.4.6 Nonstationary LSP Criteria 6023.4.7 Graphic Notation of Aggregation Structures 6063.5 Canonical Aggregation Structures 6113.5.1 Conjunctive CAS with Increasing Andness 6113.5.2 Disjunctive CAS with Increasing Orness 6143.5.3 Aggregated Mandatory/Optional and Sufficient/Optional CAS 6163.5.4 Design of a Simple LSP Evaluator Tool 6173.5.5 Distributed Mandatory/Optional and Sufficient/Optional CAS 6193.5.6 Nested Mandatory/Desired/Optional and Sufficient/Desired/Optional CAS 6213.5.7 Decreasing Andness and Decreasing Orness CAS 6223.6 Cost/Suitability Analysis as a Graded Logic Problem 6233.6.1 Cost Analysis 6233.6.2 Cost/Suitability Analysis Based on Linear Equi-Value Model 6263.6.3 Using Cost/Suitability Analysis in Competitive Bidding 6273.6.4 Conjunctive Suitability-Affordability Method 6303.7 Sensitivity Analysis and Tradeoff Analysis 6353.7.1 Sensitivity Analysis 6353.7.1.1 Sensitivity with Respect to Input Suitability Scores 6373.7.1.2 Sensitivity Properties of Basic Aggregators 6413.7.1.3 Sensitivity with Respect to Input Attributes 6433.7.2 Tradeoff Analysis 6443.7.2.1 Compensatory Properties of LSP Criteria and Graded Logic Aggregators 6473.7.2.2 The Concept of Compensation Ratio 6513.8 Reliability Analysis 6553.8.1 Sources of Errors in LSP Criteria and Their Empirical Analysis 6553.8.2 The Problem of Confidence in Evaluation Results 6603.8.3 Case Study of Reliability Analysis for a Computer Evaluation Project 6643.9 System Optimization 6713.9.1 Three Fundamental Constrained Optimization Problems 6713.9.2 The Cloud Diagram and the Set of Optimum Configurations 6733.9.3 A Case Study of Computer Configuration Optimization 6753.10 LSP Software Technology 683Part Four APPLICATIONS 6894.1 Job Selection 6934.1.1 Job Selection Attribute Tree 6944.1.2 Elementary Attribute Criteria for Job Selection 6974.1.3 Logic Aggregation of Suitability for the Job Selection Criterion 7014.1.4 A Job Selection Example 7054.2 Home Selection 7114.2.1 Home Selection Using ORE Websites and LSPhome 7114.2.2 Home Attribute Tree and Elementary Criteria 7164.2.3 Home Suitability Aggregation Structure as a Shade Diagram 7174.2.4 Using Missingness-Tolerant LSP Criteria 7254.2.5 The Optimum Home Pricing Problem 7284.2.6 A Personalized Home Selection Criterion 7314.3 Evaluation of Medical Conditions 7374.3.1 Evaluation of Disease Severity and Patient Disability 7384.3.2 Limitations of Medical Rating Scales 7404.3.3 LSP Models for Computing OSD, ODD, and PDD 7434.3.4 Evaluation of PDD for Peripheral Neuropathy 7454.3.5 The Risky Therapy Decision Problem 7524.3.6 A Case Study of Anti-MAG Neuropathy 7554.3.7 LSPmed--An Internet Tool for Medical Evaluation 7584.3.7.1 LSPmed User Types and Their Functions 7584.3.7.2 The Use of LSPmed 7604.3.7.3 Serving a Patient 7624.4 LSP Criteria in Ecology: Selecting Multi-Species Habitat Mitigation Projects 7694.4.1 Multi-Species Compensatory Mitigation Projects 7694.4.2 A Generic LSP Attribute Tree for Evaluation of Habitat Mitigation Projects 7714.4.3 Attribute Criteria and the Logic Aggregation Structure 7724.4.4 Sensitivity Analysis 7774.4.5 Logic Refining of the Aggregation Structure 7794.4.6 Cost/Suitability Analysis 7814.4.7 MSHCP Software Support 7834.5 Space Management Decision Problems 7854.5.1 A Decision Model for School Location 7854.5.1.1 Statement of the Problem 7854.5.1.2 School Locations Attribute Tree 7864.5.1.3 Elementary Criteria 7864.5.1.4 Aggregation of Suitability Degrees 7924.5.1.5 Cost Analysis 7944.5.1.6 Competitive Locations 7954.5.1.7 Cost/Suitability Analysis 7964.5.2 Suitability of Locations for Residential Development 7984.6 LSP Suitability Maps 8034.6.1 The Concept of Map Logic and LSP Suitability Maps 8034.6.2 Suitability Maps Based on Points of Interest 8064.6.3 The Problem of Optimum Location of City Objects 8104.6.4 Suitability Analysis of Urban Locations Using the LSPmap Tool 8164.6.5 GIS-LSP Suitability Maps Based on TerrSet/Idrisi 8214.6.6 GIS-LSP Suitability Maps Based on ArcGIS 8234.7 Evaluation and Comparison of Search Engines 8334.7.1 Search Engine User and Workload Models 8344.7.2 SEben--A Search Engine Benchmarking Tool 8374.7.3 LSP Criterion for Evaluation of Search Engines 8384.7.4 Search Engine Evaluation Results 843References 847Index 871

O autorze

JOZO DUJMOVI , ScD, is a professor of computer science and former chair of the Computer Science Department at San Francisco State University, where he teaches and researches soft computing, software metrics, and computer performance evaluation. He is the author of more than 170 refereed publications and the founder and principal of SEAS, a company specializing in soft computing decision models and software support for the LSP method.

Blurb

A novel approach to decision engineering, with a verified framework for modeling human reasoning

Soft Computing Evaluation Logic provides an in–depth examination of evaluation decision problems and presents comprehensive guidance toward the use of the Logic Scoring of Preference (LSP) method in modeling complex decision criteria. Fully aligned with current developments in computational intelligence, the discussion covers the design and use of LSP criteria for evaluation and comparison in diverse areas, such as search engines, medical conditions, real estate, space management, habitat mitigation projects in ecology, and land use and residential development suitability maps, with versatile transfer to other similar decision–modeling contexts.

Human decision making is rife with fuzziness, imprecision, uncertainty, and half–truths yet humans make evaluation decisions every day. In this book, such decision processes are observed, analyzed, and modeled. The result is graded logic, a soft computing mathematical infrastructure that provides both formal logic and semantic generalizations of classical Boolean logic. Graded logic is used for logic aggregation in the context of evaluation models consistent with observable properties of human reasoning.  The LSP method, based on graded logic and logic aggregation, is a vital component of an industrial–strength decision engineering framework. Thus, the book:

• Provides detailed theoretical background for graded logic
• Provides a theory of logic aggregators
• Explains the LSP method for designing complex evaluation criteria and their use
• Shows techniques for evaluation, comparison, and selection of complex systems, as well as the cost/suitability analysis, optimization, sensitivity analysis, tradeoff analysis, and missingness–tolerant aggregation
• Includes a survey of available LSP software tools, including ISEE, ANSY and LSP.NT.

With quantitative modeling of human reasoning, novel approaches to modeling decision criteria, and a verified decision engineering framework applicable to a broad array of applications, this book is an invaluable resource for graduate students, researchers, and practitioners working within the decision engineering realm.