The geographical decision-making chain: formalization and application to maritime risk analysis
1 like522 views
The document discusses the geographical decision-making chain in maritime risk analysis, highlighting the prevalence of maritime accidents and the need for improved surveillance systems. It emphasizes the importance of utilizing spatial data mining and online analytical processing to support decision-making processes in maritime safety. The conclusion proposes a framework for integrating these techniques to enhance the analysis and identification of risky maritime behaviors.
The geographical decision-making chain: formalization and application to maritime risk analysis
- 1. The geographical decision-making chain: formalization and application to maritime risk analysis Centre de recherche sur les Risques et les Crises Bilal IDIRI, Aldo NAPOLI MINES ParisTech Centre for Research on Risk and Crisis The 6th International Workshop on Information Fusion and Geographic Information Systems (IF&GIS 2013) St Petersburg, Russia, May, 2013
- 2. • Worldwide, there are still many thousands of maritime accidents each year, • 445 acts of piracy recorded (+8.5% in one year) and 1181 marine taken hostages in 2010 (BMI, 2010), • 54 700 tonnes of oil and hazardous substances accidentally discharged in 2009 against 7500 tonnes in 2008 (Cedre, 2009) Context 2 • 90% of international trade, • 80% of energy transport, • 1.19 billion Deadweight tons (dwt) in 2009, 6.7% more compared to 2008 (CNUCED, 2009). Maritime activity: risks generator context A risk is the eventuality of event that may cause harmful effects (Boisson, 1998) • Risks related to safety: collision, grounding, etc. • Risks related to security: piracy, illicit goods, etc. The need to use means of monitoring, control and repression • Organisms responsible for safety and security, • Regulations (ENC, legislative packages Erika I, II, III, etc.), • Systems of navigational aid (NavTrack, Marine GIS, ex-Trem, etc..), • Maritime tracking systems (Spationav, SIVE, SYTAR, etc..) The importance of maritime activity Risk always important despite means implemented Introduction > Proposition > Application > Conclusion
- 3. Maritime tracking systems 3 Definition: they allow the retrieval and fusion of information on vessels (position, heading, speed, etc.) for monitoring traffic on a display device (screen, touch table, etc.) . Source (DenisGouin, 2010) Control interface Maritime surveillance operator Data acquisition infrastructure Improvement of these systems: Increase detection capabilities: new sensors (FMCW radar for small boats, waves of long range surface) Integrate new databases to analyze weather, oceanographic context, etc. Integrate the decision aid tools to post-hoc analysis and identify risk behaviors Risk behavior: movement (s) + conditions describing a risk • Movement: spatio-temporal (position change) • Conditions: vessel characteristics, environment, etc. Introduction > Proposition > Application > Conclusion
- 4. Problematic 4 The maritime surveillance systems display operational data difficult to exploit for decision-making These data are not saved to a post-hoc analysis Decision-making functions of collecting multi-source data to presentation to users are not supported The problem of maritime surveillance can be seen as a problem of spatiotemporal decision aid Operators should monitor ships as mobile objects moving in a spatiotemporal open space and make decisions on risky behaviors OperationalOperational Decision supportDecision support DataData Immediate Historical Detailed Aggregated Internal to the system Multiple sources Normalised De-normalised InterfaceInterface Complex Intuitive QueriesQueries Predefined user queries Open-ended Slow response to aggregated queries Frequent updates. Rapid response to aggregated queries No update. Introduction > Proposition > Application > Conclusion
- 5. Approaches to decision support 5 Many ways to improve decision support in the maritime domain Approaches based on advanced spatial analysis (Claramunt et al., 2007) Knowledge representation (Vandecasteele et al.,2012) Automatic risk identification (Idiri et al., 2012) Knowledge extraction about risk behavior based on historical data Introduction > Proposition > Application > Conclusion
- 6. Geo-decision-making chain 6 Spatial data mining Spatial OnLine Analytical Processing Formalize the use of geo-decision-making tools in the form of a chain (using the analogy of the decision-making chain) with the aim of support all Geo-decision-making functions Spatial data mining and Spatial OnLine Analytical Processing allows the extraction of: Patterns (spatial and temporal unusual frequent, periodic, groups, etc.) Relationships (association rules, sequential, etc.). Introduction > Proposition > Application > Conclusion
- 7. The post-hoc analysis of maritime risks 7 The data repository (or SDW) that records the evolution of maritime activity serves as a tool for automatic (SDM) data mining and visual (SOLAP) data mining that leads to knowledge discovery on risk behaviors. Definition: Spatial Data Mining is the non-trivial extraction of implicit and potentially useful knowledge from data supplied by the spatial database (Krzysztof et al., 1995) The Spatial OLAP has been defined by Bédard as, “a visual platform specifically designed to support a rapid and efficient spatiotemporal analysis through a multidimensional approach that includes mapping, graphical and tabular levels of aggregation” (Bédard , 1997) Introduction > Proposition > Application > Conclusion
- 8. Application to the maritime 8 MAIB data: Historical accidents/incidents between 1991 and 2009, 14,900 accidents and incidents, 16,230 ships. MERRA data: Provides meteorological data from the period 1991-2009, Regular download (1 time/day) to supply the facts. AIS data: Historical data since ~ 4 months, Continuous flow of ship movements. Introduction > Proposition > Application > Conclusion The databases
- 9. Spatial data mining 9 Eps 20Km ~ 0.18° MinPts 14 cas Rules linking vessel characteristics, the environment and the different types of marine accidents Rules linking vessel characteristics, the environment and the different types of marine accidents Discovery of accident-prone areasDiscovery of accident-prone areas Identification of behaviors do not like the behavior of nearby vessels Identification of behaviors do not like the behavior of nearby vessels Introduction > Proposition > Application > Conclusion
- 10. Spatial OnLine Analytical Processing 10 T. LASVENES and V. BENEDETTI, 2012 Line identification of dangerous goods by visual data miningLine identification of dangerous goods by visual data mining Introduction > Proposition > Application > Conclusion
- 11. Conclusion 11 Results Formalize a paradigm that we call geo-decision-making chain Apply the geo-decision-making chain to the domain of maritime surveillance that supports all decision-making functions, from data collection to its presentation to decision-makers Automatic and visual extraction of risky behaviors from data Introduction > Proposition > Application > Conclusion Future works Apply other data mining algorithms (periodic patterns, convoy, etc.) to extract risky behaviors Design and develop a software workshop integrating automatic methods for data mining and visual aid for analysis and subsequent identification of risk behaviors
- 12. Centre de recherche sur les Risques et les Crises Thanks for attention Personal page and publications http://www.mines-paristech.fr/Services/Annuaire/aldo-napoli Aldo NAPOLI Phone: +33 (0) 4 93 67 89 15 Fax. : +33 (0) 4 93 95 75 81 E-mail: : aldo.napoli@mines-paristech.fr ?
Editor's Notes
- #7: To achieve this integration of data mining into maritime surveillance systems, we formalized the use of geo-decision-making tools in the form of a chain (using the analogy of the decision-making chain).