Agents and Data Mining Interaction : 4th International Workshop on Agents and Data Mining Interaction, ADMI 2009, Budapest, Hungary, May 10-15,2009, Revised Selected Papers

Annotation This book constitutes the thoroughly refereed post-conference proceedings of the 4th International Workshop on Agents and Data Mining Interaction, ADMI 2009, held in Budapest, Hungary in May 10-15, 2009 as an associated event of AAMAS 2009, the 8th International Joint Conference on Autonomous Agents and Multiagent Systems. The 12 revised papers and 2 invited talks presented were carefully reviewed and selected from numerous submissions. Organized in topical sections on agent-driven data mining, data mining driven agents, and agent mining applications, the papers show the exploiting of agent-driven data mining and the resolving of critical data mining problems in theory and practice; how to improve data mining-driven agents, and how data mining can strengthen agent intelligence in research and practical applications. Subjects that are also addressed are exploring the integration of agents and data mining towards a super-intelligent information processing and systems, and identifying challenges and directions for future research on the synergy between agents and data mining Introduction......Page 11 Background......Page 12 Bioinformatics Databases......Page 14 An Agent-Based Environment for Automating Annotation......Page 15 Integrating Knowledge through Cooperative Negotiation......Page 19 Case Study in Automated Annotation of Proteins......Page 20 Selection of Data Sets of Motifs as Attributes in the Process of Automating the Annotation of Proteins' Keywords......Page 21 Enzyme Classification......Page 23 Multiagent Clustering......Page 24 Conclusion......Page 27 Extended Abstract......Page 29 Introduction......Page 31 What Is Data Intelligence......Page 32 Aspects of Data Intelligence......Page 33 Aspects of Domain Intelligence......Page 34 Aims of Involving Network Intelligence......Page 35 What Is Human Intelligence......Page 36 Aspects of Human Intelligence......Page 37 Aspects of Organizational Intelligence......Page 38 Aims of Involving Social Intelligence......Page 39 Aspects of Social Intelligence......Page 40 Conclusion......Page 41 Introduction......Page 44 Structure of the System......Page 45 System Functioning Algorithm......Page 47 Gathered Results......Page 54 Conclusions......Page 56 Motivation and Goals......Page 58 Related Work......Page 60 System Structure......Page 61 Agent Interactions......Page 63 System Extendibility......Page 65 System Demonstration......Page 66 Meta ARM (Association Rule Mining) Scenario......Page 67 Classifier Generation Scenario......Page 69 Conclusions......Page 71 Introduction......Page 74 Particle Swarm Optimization Exposed......Page 75 Bacterial Foraging Algorithms......Page 77 PSO Clustering......Page 78 Automatic PSO Clustering......Page 79 AutoCB Clustering......Page 80 The AutoCPB Algorithm......Page 82 Experimental Results......Page 83 Cluster Validation......Page 84 Benchmark Testing......Page 85 Final Discussion......Page 87 Introduction......Page 89 Overview of the Proposed Multiagent System......Page 90 New Growing SOM Model......Page 91 Experimental Results......Page 94 Conclusions......Page 97 Introduction......Page 100 Related Work......Page 101 Relevant Measures......Page 102 Private Knowledge Loss......Page 103 Classification Accuracy......Page 105 Joining Naïve Bayes Models......Page 106 Model and Sub-model Sharing......Page 107 Experiment Domain and Setting......Page 108 Results for the Model Sharing Strategy......Page 109 Results for the Data Classification Strategy......Page 111 Conclusion......Page 112 Introduction......Page 114 Power Markets......Page 115 Power Market Auctions......Page 116 State-of-the-Art......Page 117 Cassandra Architecture......Page 118 Cassandra Users......Page 120 Implementation......Page 121 Preliminary Results......Page 122 Training......Page 123 Meta Classification......Page 124 Pilot Case Scenario......Page 125 Future Work......Page 126 Introduction......Page 129 Decision Support......Page 130 Multiagent System......Page 131 Methodology......Page 132 MAS Architecture......Page 133 DM Models......Page 135 DM Results......Page 136 Conclusion......Page 139 Introduction......Page 142 Zero-Intelligence Plus (ZIP)......Page 143 Reverse TAC (``CAT'') Game......Page 144 Market Client......Page 146 Pricing Policy......Page 147 Modeling the Problem......Page 148 Our Approach......Page 149 Related Work......Page 151 Conclusions and Future Work......Page 152 Introduction......Page 155 Overview......Page 158 Overview of Agent Assignment Strategies......Page 159 PAPE System Architecture......Page 160 Preprocessing Phase......Page 161 Integrated Data Structure Generation Phase......Page 162 Pattern Based Agent Performance Evaluation......Page 163 Agent Assignment Learning Phase......Page 164 Agent Assignment Updating Phase......Page 165 Experimental Results......Page 166 Related Work......Page 167 Conclusion and Future Work......Page 168 Introduction......Page 170 Representing Knowledge......Page 172 Actively Learning a Concept......Page 174 Selection of Examples......Page 175 Significance of Instances......Page 176 Putting It Together......Page 177 Evaluation......Page 178 Discussion......Page 181 Introduction......Page 183 Complex Systems Analysis Applied to the Web and Economics......Page 184 Influence of Display Rank on Clicking Behavior......Page 185 Results on Display Position Bias and Interpretation......Page 186 Market Structure at the Advertiser Level......Page 187 Distribution of Impressions vs. Distribution of Clicks for the Top Advertisers......Page 188 Distribution of Market Share Per Display Rank Position......Page 189 Using Click Data to Derive Search Term Recommendations......Page 190 Constructing Keyword Correlation Graphs......Page 191 Graph Correlation Graphs: Results......Page 192 Automatic Identification of Sets of Keywords......Page 193 The Graph Partitioning Algorithm......Page 194 Discussion of Graph Partitioning Results......Page 195 Future Work......Page 196 The2009InternationalWorkshoponAgentsandDataMiningInteraction(ADMI 2009) was a joint event with AAMAS2009. In recentyears,agents and data mining interaction (ADMI), or agent mining forshort,hasemergedasaverypromisingresearch?eld. Followingthesuccessof ADMI 2006 in Hong Kong, ADMI 2007 in San Jose, and ADMI 2008 in Sydney, the ADMI 2009 workshop in Budapest provided a premier forum for sharing research and engineering results, as well as potential challenges and prospects encountered in the synergy between agents and data mining. As usual, the ADMI workshop encouraged and promoted theoretical and applied research and development, which aims at: – Exploitingagent-drivendatamininganddemonstratinghowintelligentagent technology can contribute to critical data mining problems in theory and practice – Improving data mining-driven agents and showing how data mining can strengthen agent intelligence in research and practical applications – Exploring the integration of agents and data mining toward a super-intelligent information processing and systems – Identifying challenges and directions for future research on the synergy between agents and data mining ADMI 2009 featured two invited talks and twelve selected papers. The?rst invited talk was on “Agents and Data Mining in Bioinformatics,” with the s- ond focusing on “Knowledge-Based Reinforcement Learning. ” The ten accepted papers are from seven countries. A majority of submissions came from Eu- pean countries, indicating the boom of ADMI research in Europe. In addition the two invited papers, addressed fundamental issues related to agent-driven data mining, data mining-driven agents, and agent mining applications. The proceedings of the ADMI workshops will be published as part of the LNAIseriesbySpringer.WeappreciatethesupportofSpringer,andinparticular Alfred Hofmann. The2009InternationalWorkshoponAgentsandDataMiningInteraction(ADMI 2009) was a joint event with AAMAS2009. In recentyears,agents and data mining interaction (ADMI), or agent mining forshort,hasemergedasaverypromisingresearch?eld. Followingthesuccessof ADMI 2006 in Hong Kong, ADMI 2007 in San Jose, and ADMI 2008 in Sydney, the ADMI 2009 workshop in Budapest provided a premier forum for sharing research and engineering results, as well as potential challenges and prospects encountered in the synergy between agents and data mining. As usual, the ADMI workshop encouraged and promoted theoretical and applied research and development, which aims Exploitingagent-drivendatamininganddemonstratinghowintelligentagent technology can contribute to critical data mining problems in theory and practice Improving data mining-driven agents and showing how data mining can strengthen agent intelligence in research and practical applications Exploring the integration of agents and data mining toward a super-intelligent information processing and systems Identifying challenges and directions for future research on the synergy between agents and data mining ADMI 2009 featured two invited talks and twelve selected papers. The ?rst invited talk was on Agents and Data Mining in Bioinformatics, with the s- ond focusing on Knowledge-Based Reinforcement Learning. The ten accepted papers are from seven countries. A majority of submissions came from Eu- pean countries, indicating the boom of ADMI research in Europe. In addition the two invited papers, addressed fundamental issues related to agent-driven data mining, data mining-driven agents, and agent mining applications. The proceedings of the ADMI workshops will be published as part of the LNAIseriesbySpringer. WeappreciatethesupportofSpringer,andinparticular Alfred Hofmann.