Research and Development in Intelligent Systems XXVII (eBook)

ACKNOWLEDGEMENT 6
CONTENTS 11
Research and Development inIntelligent Systems XXVII 16
BEST TECHNICAL PAPER 17
Effective Product Recommendation Using theReal-TimeWeb 18
Abstract 18
1 Introduction 18
2 Related Work 20
3 The Blippr Service 21
4 Product Recommendation using RTW Data 22
4.1 Index Creation 23
4.2 Recommending Products 24
5 Evaluation 24
5.1 Datasets 25
5.2 Metrics 26
5.3 Recommendation Results 26
6 Conclusions 29
References 30
INTELLIGENT AGENTS 32
Agent Argumentation with Opinions and Advice 33
Abstract 33
1 Introduction 33
2 Opinions and Advice 34
3 Relationships 36
4 The Relationship Model 38
4.1 The Intimacy Model: 40
4.2 The Reliability Model: 42
5 Trust and Integrity 43
6 ‘Relationship-aware’ Argumentation Strategies 44
7 Discussion 45
References 46
Graph-Based Norm Explanation 47
Abstract 47
1 Introduction 48
2 Background 49
2.1 The Normative Model 49
2.2 Conceptual Graphs 52
3 Graphically Computing the Status of Norms 54
3.1 Modelling Norms with CGs 54
3.2 Instantiating Norms 57
3.3 Computing the Status of Norms 57
4 Discussion 58
References 60
Modelling Social Structures and Hierarchies inLanguage Evolution 61
Abstract 61
1 Introduction 61
2 Modelling Approach 62
3 Model Implementation 64
4 Experiment Design 66
5 Results and Discussion 68
6 Conclusions 72
References 73
KNOWLEDGE DISCOVERY AND DATA MINING 75
On the Usefulness ofWeight-Based Constraints in Frequent Subgraph Mining 76
Abstract 76
1 Introduction 76
2 Preliminaries 78
3 Weight-Based Constraints 80
4 Weight-Based Mining 81
5 Weighted Graph Mining Applied 83
6 Experimental Evaluation 84
7 Related Work 87
8 Conclusions 88
References 89
Induction of Modular Classification Rules:Using Jmax-pruning 90
Abstract 90
1 Introduction 90
2 The Prism Family of Algorithms 91
2.1 Dealing with Clashes 93
2.2 Dealing with Continuous Attributes 93
2.3 J-pruning 94
3 Variation of J-pruning 95
3.1 Critique of J-pruning 95
3.2 Jmax-pruning 95
4 Evaluation of Jmax-pruning 97
5 Ongoing Work 99
5.1 J-PrismTCS 99
5.2 Jmax-Pruning for TDIDT 99
6 Conclusions 102
References 102
A Kolmogorov Complexity View of Analogy:From Logical Modeling to Experimentations 104
Abstract 104
1 Introduction 104
2 Analogical proportions: a logical view 106
2.1 Brief analysis 106
2.2 Formal setting 107
2.3 Boolean interpretation 108
2.4 Formal frameworks to cope with natural language analogies 109
3 Analogies in natural language: an information-theoretic view 110
3.1 Kolmogorov theory: a brief overview 110
4 Kolmogorov model for analogy in natural language 111
5 Experimentations 112
5.1 Probability distribution generator 112
5.2 Results 113
5.3 A new option 114
6 Related works 115
7 Conclusion 116
References 117
Evolving Temporal Association Rules withGenetic Algorithms 118
Abstract 118
1 Introduction 118
2 Related Work 119
2.1 Temporal Association Rule Mining 119
2.2 Association Rule Mining with Evolutionary Computation 121
3 Evolving Temporal Association Rules 122
4 Evaluation 125
4.1 Methodology and Datasets 126
4.2 Results 127
5 Conclusion 129
References 130
PLANNING AND SCHEDULING 132
PIPSS*: A System Based on Temporal Estimates 133
Abstract 133
1 Introduction 133
2 PIPSS 134
2.1 HPP: Heuristic Progressive Planner 134
2.2 ISES Algorithm 135
2.3 PIPSS Architecture 135
3 PIPSS* 136
3.1 Relaxed GraphPlan 136
3.2 Extension 137
4 Experimental Results 141
4.1 Satellite Domain 142
4.2 PipesWorld Domain 143
4.3 Openstacks Domain 144
5 Conclusions 145
References 146
Extending SATPLAN to Multiple Agents 147
Abstract 147
1 Introduction 147
2 Propositional Satisfiability based Planning:SATPLAN 149
3 MACAP and MAOCAP 149
4 Solving the MACAP using µ-SATPLAN 150
4.1 Independent Plan Computation 151
4.2 Coordinated Plan Computation 152
4.2.1 Handling Negative Interactions 152
4.2.2 Handling Positive Interactions and Parallel Actions Interference 153
4.3 Coordinated Plan for nth agent 155
5 Solving the MAOCAP 155
6 Experimental Results 157
7 Conclusion and future work 159
References 160
MACHINE LEARNING 161
A New Approach for Partitional ClusteringUsing Entropy Notation and Hopfield Network 162
Abstract 162
1 Introduction 162
2 SOCHOM Method 163
3 Algorithm Descriptions 167
4 Experimental results 170
5 Conclusions 171
References 171
Hierarchical Traces for ReducedNSM Memory Requirements 173
Abstract 173
1 Introduction and Motivation 173
2 Related Work 174
3 NSM with Hierarchical Traces 175
3.1 Hierarchical Traces 176
3.2 Estimating Discounted Rewards 178
3.3 Hidden State Identification 179
4 Re-use and Sub-Sequence Length 180
4.1 Speed of Learning 181
5 Memory Persistence 183
6 Conclusions and FutureWork 185
References 185
On Reinforcement Memory for Non-MarkovianControl 187
Abstract 187
1 Introduction 187
2 A non-Markovian and Perceptual Aliasing 189
2.1 POMDP Formal Setting 189
3 Self-Optimizing Controller Architecture 190
4 Non-Markovian Memory Controller 192
4.1 Conventional Memory Controller 192
4.2 Self-Optimizing Memory Controller 192
5 Memory-Capable Function Approximation 192
5.1 Actor-Critic Learning 192
5.2 AC in non-Markovian Domain 193
5.3 Decision-tree Ensemble Memory for Optimal Learning 194
6 Experiment and Results 194
6.1 Related work 195
6.2 Non-Markovian Cart Pole Balancing 195
6.3 Non-Markovian Two-Pole Balancing 196
7 Conclusions 197
References 198
Appendix 199
8 Pole-balancing learning parameters 199
9 Parameters for comparisons in cart pole balancing 200
A Fast Approximated Evolutionary Approach toImprove SVM Accuracy 201
Abstract 201
1 Introduction 201
2 Feature and Model Selection 202
2.1 Problem Overview 202
3 GA-based Method 203
3.1 Genetic Algorithms and Support Vector Machines 203
3.1.1 Genetic Algorithms 204
3.2 A Kernel Matrix-based Approach 205
3.2.1 SVMs Optimization Problem 205
3.2.2 Performance’s Estimation Through Kernel Matrix 206
3.2.3 KernelMatrix Criteria 207
3.2.4 Error’sMeasures for Kernel Matrix Criteria 208
4 Experiments 208
4.1 Experiments – Behavior on Training Set 210
4.2 Experiments – Classifiers’ Performances on Test Set 210
5 Conclusion 213
References 213
EVOLUTIONARY ALGORITHMS, BAYESIANNETWORKS AND MODEL-BASED DIAGNOSIS 215
A Particle Swarm Optimization Approach forthe Case Retrieval Stage in CBR 216
Abstract 216
1 Introduction 216
2 Case Retrieval Stage 217
2.1 Theoretical Background 217
2.2 Proposed Architecture 219
3 The PSO Approach 220
3.1 PSO Background 221
3.2 Proposed Approach 223
4 Experimental Results and Discussion 225
5 Conclusion and future work 228
References 229
Dynamic Pricing with Neural Network DemandModels and Evolutionary Algorithms 230
Abstract 230
1 Introduction 230
2 A Mathematical Model of Dynamic Pricing 231
3 Model of demand 233
3.1 Estimating parameters of the demand model 233
4 Neural network demand models 234
4.1 Estimating parameters of the neural networks 236
5 An EA approach to dynamic pricing 236
5.1 Constraint handling in EA 236
5.2 Solution representation in EA 237
6 Experiments and results 237
6.1 Experimental setups 238
6.2 Results 239
7 Conclusions 242
References 243
Discretisation Does Affect the Performance ofBayesian Networks 244
Abstract 244
1 Introduction 244
2 Materials and Methods 246
2.1 Bayesian Networks 246
2.2 Discretisation 246
2.2.1 Previous Research 247
2.2.2 Discretisation Methods 247
3 Discretisation in Automated Mammographic Analysis 249
3.1 Mammographic Analysis 249
3.2 Bayesian Network Model 250
3.3 Data and Experimental Set-up 252
3.4 Results 253
4 Conclusions 255
References 257
A Structural Approach to Sensor Placementbased on Symbolic Compilation of the Model 258
Abstract 258
1 Introduction 258
2 Discriminability and MASS for Qualitative Relational Models 259
3 Background Information on Numerical Models 261
4 Building a Qualitative Relational Model 263
5 Computation of MASS 267
6 Application to the GFS 269
7 Conclusions 270
References 271
SHORT PAPERS 272
Artificial Immunity Based CooperativeSustainment Framework for Multi-AgentSystems 273
Abstract 273
1 Introduction 273
2 AIS-based Cooperative Sustainment Control Framework 274
2.1 Overview on the Control Architecture 274
2.2 Sustainment Operations 275
3 Experiment and Result 276
4 Conclusion and Future Works 277
References 278
The Mining and Analysis Continuumof Explaining Uncovered 279
Abstract 279
1 Introduction 279
2 Explanation-Aware Software Design and Computing 280
3 Goals and Kinds of Explanations 281
4 The Mining and Analysis Continuum of Explaining 281
4.1 Explanation-Aware Mining and Analysis 281
4.2 Explanation Dimensions (Continuum) 282
5 Summary and Outlook 284
References 284
Genetic Folding: A New Class of EvolutionaryAlgorithms 285
Abstract 285
1 Introduction 285
2 Problem Definition 286
3 Genetic Folding Algorithm 286
3.1 Genetic Folding Programming 287
3.2 Encoding and Decoding Procedures 287
4 Experimental Design and Results 288
5 Conclusions 290
References 290
SOMA: A Proposed Framework for Trend Mining in Large UK Diabetic RetinopathyTemporal Databases 291
Abstract 291
1 Introduction 291
2 Diabetic Retinopathy Databases 292
3 The SOMA Trend Mining Framework 292
4 Experimental Evaluation 294
5 Conclusion 296
References 296
Applications and Innovationsin Intelligent Systems XVIII 297
BEST APPLICATION PAPER 298
Artificial Intelligence Techniques for the BerthAllocation and Container Stacking Problems inContainer Terminals 299
Abstract 299
1 Introduction 299
2 An Integrated Approach for Container Stacking and BerthAllocation Problems 301
3 A Domain-dependent Planner for the Container StackingProblem 303
4 The Berth Allocation Problem 305
4.1 A meta-heuristic method for BAP 306
5 Evaluation 309
6 Conclusions 311
Acknowledgments 312
References 312
APPLICATIONS OF MACHINE LEARNING I 313
Social Network Trend Analysis Using FrequentPattern Mining and Self Organizing Maps 314
Abstract 314
1 Introduction 314
2 PreviousWork 316
3 The Trend Mining Mechanism 317
3.1 Frequent Pattern Trend Mining 317
3.2 Trend Clustering 319
3.3 Analysis of Trend Clusters 320
4 Experimental Analysis Using The Cattle Movement SocialNetwork 321
4.1 Cattle Movement Database 321
4.2 Cattle Movement Trend Mining 322
5 Car Insurance Trend Mining 324
6 Conclusion 326
References 327
Retinal Image Classification for the Screening ofAge-Related Macular Degeneration 328
Abstract 328
1 Introduction 329
2 Age-related Macular Degeneration 330
3 PreviousWork 331
4 The AMD Screening Process 332
5 Image Pre-processing 333
5.1 Image Enhancement 333
5.2 Objects Segmentation 334
6 Spatial Histogram Generation 335
7 Feature Selection 336
8 Retinal Image Classification using CBR and DTW 337
9 Evaluation 337
9.1 Number of Bins Parameter 338
9.2 T Parameter Identification 338
10 Conclusion 340
References 340
An Ensemble Dynamic Time Warping Classifierwith Application to Activity Recognition 342
Abstract 342
1 Introduction 342
2 Methodology 343
2.1 Dynamic Time Warping 343
2.2 Combining DTW Similarity Scores for Classification 343
3 Key Issues 346
3.1 Sliding Window & Warping Window Size
3.2 Similarities in Activities 346
3.3 Transitioning between Activities 347
4 Experiments 348
4.1 Description of Data 348
4.2 Training Procedure 348
4.3 Testing Procedure 348
5 Results 349
5.1 Using Individual Sensors 349
5.2 Using All Three Sensors 351
6 Related Research 353
7 Conclusions 354
References 354
APPLICATIONS OF MACHINE LEARNING II 356
Self-Adaptive Stepsize Search Applied toOptimal Structural Design 357
Abstract 357
1 Introduction 357
2 The 25-bar Problem 358
3 Self-Adaptive Stepsize Search 362
4 Experimental Results 363
5 Conclusions 365
References 366
Health Problems Discovery from Motion-Capture Data of Elderly 367
Abstract 367
1 Introduction 367
2 Related Work 368
3 Methods and Materials 371
3.1 Targeted Health Problems 371
3.2 Construction of the Features for Machine Learning 371
3.3 Modeling Target Health Problems using Machine Learning 372
4 Experiments and results 373
4.1 Variation of Noise 374
4.2 Reduction of the Number of Tags 375
4.3 Dependence of the Classification Accuracy on the TagPlacement and Noise Level 376
5 Explanation of the Detection 377
6 Conclusion 378
Acknowledgments 379
References 379
Selecting Features in Origin Analysis 381
Abstract 381
1 Introduction 381
2 Application 382
3 Data Collection 383
3.1 Collection and Preprocessing 383
3.2 Filtering 383
3.3 Experimental Data 385
4 Feature Construction 386
4.1 Tools 386
4.2 Measurements 387
4.2.1 Ferret and Duplo Basics 387
4.2.2 Ferret Trigrams 388
4.2.3 Block Based Measures 388
4.3 The Feature Set 389
5 Experiments 390
5.1 Method 390
5.2 Results 390
6 Discussion and Summary 391
References 394
AI FOR SCHEDULING 395
An Extended Deterministic Dendritic CellAlgorithm for Dynamic Job Shop Scheduling 396
Abstract 396
1 Introduction 396
2 Dynamic JSSP Definition 398
3 The Extended Deterministic Dendritic Cell Algorithm 399
3.1 The Principles of DCA 399
3.2 The Extended Deterministic DCA (dDCA) 400
4 Extended dDCA for Dynamic JSSP 401
5 Experiments and Results 405
6 Conclusions 408
References 409
Reinforcement Learning for Scheduling ofMaintenance 410
Abstract 410
1 Introduction 410
2 Reinforcement Learning 411
3 Problem Formulation 412
3.1 Plant model 413
3.2 Reinforcement Learning Model 414
3.3 Experiments 415
4 Results 415
4.1 Level 1: Basic Model 415
4.2 Level 2: Condition Data 417
4.3 Level 3: Energy Consumption Data 418
4.4 Level 4: Complex System 419
5 Discussion and Conclusions 420
References 421
AI IN ACTION 424
Genetic Evolution and Adaptation of AdvancedProtocols for Ad Hoc Network HardwareSystems 425
Abstract 425
1 Introduction 425
2 Background 426
2.1 Protocol stack 426
2.2 Related Research 427
3 Protocol Methodology 427
3.1 General Concept 427
3.2 Alphabet of Characteristics 428
3.3 Physical Layer Characteristics 428
3.4 Media Access Control Layer Characteristics 429
3.5 Network Layer Routing Characteristics 430
3.6 Interfacing Sub-Protocols 431
3.7 The Genetic Algorithm and Fitness Function 431
4 Network Scenario 432
5 Results 434
6 Conclusion 436
Acknowledgments 437
References 437
The Next Generation of Legal Expert Systems -New Dawn or False Dawn? 439
Abstract 439
1 Introduction 439
2 The Jaes Project 440
3 What are the particular challenges of the legal domain? 441
3.1 Complexity 441
3.2 Uncertainty 442
3.3 Financial Disincentives 442
4 Alternatives for the next generation of legal expert systems 442
4.1 Case Based Reasoning (CBR) 443
4.2 Blackboard architecture 444
4.3 Service-oriented architecture (SOA) 446
4.4 Hybrid integration 448
5 Will the next generation of systems overcome the challenges? 450
References 451
Incorporating Semantics into Data DrivenWorkflows for Content Based Analysis 453
Abstract 453
1 Introduction 453
2 Linguistic Analysis Using NLP Oriented Approaches 454
3 Sentiment Analysis 456
4 Semantically Meaningful Analysis 457
4.1 Using Ontologies as Lexical Resources 457
4.2 Using SNOMED CT as Lexical Resource 458
4.3 Using UMLS to enable Semantically Meaningful Analysis 458
5 Content Based Analysis Results 461
6 Semantic Annotation of Data and Functionality 462
7 Concluding Remarks 465
References 465
GhostWriter-2.0: Product Reviews withCase-Based Support 467
Abstract 467
1 Introduction 467
2 An End-User View of GhostWriter-2.0 469
3 Cases in GhostWriter-2.0 470
4 How Ghostwriter-2.0 Works 471
4.1 Populating the case base 472
4.2 Making suggestions 472
4.2.1 Retrieval 473
4.2.2 Suggestion selection 473
5 Experimental Evaluation 475
6 Related Research 478
7 Conclusions and Future Work 479
References 480
SHORT PAPERS 481
Dynamic Programming Algorithm vs. GeneticAlgorithm: Which is Faster? 482
Abstract 482
1 Introduction 482
1.1 Related Work 483
2 Implementation of GA in PostgreSQL 484
3 Experimental Evaluations 484
3.1 Evaluation with MS SQL Server 485
3.2 Comparison of GEQO Module with the PostgreSQL DynamicProgramming Component 486
4 Conclusions 486
References 487
Automatic Detection of Pectoral Muscle with theMaximum Intensity Change Algorithm 488
Abstract 488
1 Introduction 488
2 Pectoral Muscle Detection 489
2.1 Detect the first point of edge of pectoral muscle 489
2.2 Detect the cliff of pectoral muscle 490
2.3 Enclose the pectoral muscle area 491
3 Experimentation and Results 492
4 Conclusion 493
References 493