Behavior Informatics and Analytics: Let Behavior Talk Longbing Cao Data Sciences & Knowledge Discovery Lab Centre for Quantum Computation and Intelligent Systems University of Technology, Sydney, Australia

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Outline Motivation Behavior and Behavioral Model BIA Framework BIA Theoretical Underpinnings BIA Research Issues BIA Applications & Case Studies BIA References

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Motivation Behavior is an important analysis object in Business intelligence Customer relationship management Social computing Intrusion detection Fraud detection Event analysis Market strategy design Group decision-making, etc.

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Examples Customer behavior analysis Consumer behavior and market strategy Web usage and user preference analysis Exceptional behavior analysis of terrorist and criminals Trading pattern analysis of investors in capital markets

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Traditional analysis on behavior Behavior-oriented analysis was usually conducted on customer demographic and transactional data directly Telecom churn analysis, customer demographic data and service usage data are analyzed to classify customers into loyal and non-loyal groups based on the dynamics of usage change outlier mining of trading behavior, price movement is usually focused to detect abnormal behavior so-called behavior-oriented analysis is actually not on customer behavior-oriented elements, rather on straightforward customer demographic data and business usage related appearance data (transactions)

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Market price trend/movement estimation

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Problems with traditional behavior analysis customer demographic and transactional data is not organized in terms of behavior but entity relationships human behavior is implicit in normal transactional data: behavior implication cannot support in-depth analysis on behavior interior: behavior exterior Cannot scrutinize behavioral intention and impact on business appearance and problems Such behavior implication indicates the limitation or even ineffectiveness of supporting behavior-oriented analysis on transactional data directly.

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM behavior can make difference behavior plays the role as internal driving forces or causes for business appearance and problems complement traditional pattern analysis solely relying on demographic and transactional data Disclose extra information and relationship between behavior and target business problem-solving A multiple-dimensional viewpoint and solution may exist that can uncover problem-solving evidence from not only demographic and transactional but behavioral (including intentional, social and impact aspects) perspectives

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM support genuine behavior analysis make behavior ‘explicit’ by squeezing out behavior elements hidden in transactional data a conversion from transactional space to behavior feature space is necessary behavior data: behavior modeling and mapping organized in terms of behavior, behavior relationship and impact Explicitly and more effectively analyze behavior patterns and behavior impacts than on transactional data

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM main goals and tasks of behavior informatics and analytics (BIA) behavioral data construction behavior modeling and representation, behavior impact modeling, Behavior pattern analysis, and behavior presentation BIA is mainly from the perspectives of information technology and data analysis rather than from social behavior aspect

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM BIA makes difference Case study: churn analysis of mobile customers analysis on demographic and service usage data behavior sequences of a customer activities happened from his/her registration and activation of a new account into a network Characteristics of making payments to the date leaving the network  Know deep knowledge about mobile service retainer’s intention, activity change, usage dynamics, and payment profile  disclosing reasons and drivers of churners and their loyalty change

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM So, what is behavior Under the scope of Behavior Informatics and Analytics, behavior refers to those activities that present as actions, operations or events, and activity sequences conducted by human beings under certain context and environment, as well as behavior surroundings. the informatics and analytics for symbolic behavior and the analytics of mapped behavior. symbolic behavior Those social activities recorded into computer systems, which present as symbols representing human interaction and operation with a particular object or object system; place an order game user behavior intelligent agent behavior;

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM mapped behavior direct or indirect mapping of physical behavior in a virtual world. Those physical activities recorded by sensors into computer systems, human activities captured by video surveillance systems; robot’s behavior organism’s behavior in game systems;

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM An Abstract Behavioral Model Behavior attributes and properties: Subject (s): The entity (or entities) that issues the activity or activity sequence; Object (o): The entity (or entities) on which a behavior is imposed on; Context (e): The environment Goal (g): Goal represents the objectives Belief (b): Belief represents the informational state and knowledge Action (a): Action represents what the behavior subject has chosen to do or operate; Plan (l): Plans are sequences of actions Impact (f): The results led by the execution of a behavior on its object or context; Constraint (c): Constraint represents what conditions are taken on the behavior; constraints are instantiated into specific factors in a domain; Time (t): When the behavior occurs; Place (w): Where the behavior happens; Status (u): The stage where a behavior is currently located; Associate (m): Other behavior instances or sequences of actions that are associated with the target one;

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM An abstract behavior model Demographics of behavioral subjects and objects Associates of a behavior may form into certain behavior sequences or network; Social behavioral network consists of sequences of behaviors that are organized in terms of certain social relationships or norms.

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM behavior instance: behavior vector basic properties social and organizational factors vector-based behavior sequences, vector-oriented patterns.

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM vector-oriented behavior pattern analysis is much more comprehensive Behavior performer: Subject (s), action (a), time (t), place (w) Social information: Object (o), context (e), constraints (c), associations (m) Intentional information: Subject’s: goal (g), belief (b), plan (l) Behavior performance: Impact (f), status (u)  New methods for vector-based behavior pattern analysis

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM The concept of BIA BIA aims to develop methodologies, techniques and practical tools for representing, modeling, analyzing, understanding and/or utilizing symbolic and/or mapped behavior, behavioral interaction and network, behavioral patterns, behavioral impacts, the formation of behavior-oriented groups and collective intelligence, and behavioral intelligence emergence.

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Research map of BIA

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM BIA research issues Behavioral data Behavioral elements hidden or dispersed in transactional data behavioral feature space  Behavioral data modeling  Behavioral feature space  Mapping from transactional to behavioral data  Behavioral data processing  Behavioral data transformation

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Behavioral representation (behavioral modeling) describing behavioral elements and the relationships amongst the elements presentation and construction of behavioral sequences unified mechanism for describing and presenting behavioral elements, behavioral impact and patterns

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM  Behavior model  Behavior interaction  Collective behavior  Action selection  Behavior convergence and divergence  Behavior representation  Behavioral language  Behavior dynamics  Behavioral sequencing

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Behavioral impact analysis Behavioral instances that are associated with high impact on business processes and/or outcomes modeling of behavioral impact

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM  Behavior impact analysis  Behavioral measurement  Organizational/social impact analysis  Risk, cost and trust analysis  Scenario analysis  Cause-effect analysis  Exception/outlier analysis and use  Impact transfer patterns  Opportunity analysis and use  Detection, prediction, intervention and prevention

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Behavioral pattern analysis behavioral patterns without the consideration of behavioral impact, analyze the relationships between behavior sequences and particular types of impact

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM  Emergent behavioral structures  Behavior semantic relationship  Behavior stream mining  Dynamic behavior pattern analysis  Dynamic behavior impact analysis  Visual behavior pattern analysis  Detection, prediction and prevention  Customer behavior analysis  Behavior tracking  Demographic-behavioral combined pattern analysis  Cross-source behavior analysis  Correlation analysis  Social networking behavior  Linkage analysis  Evolution and emergence  Behavior clustering  Behavior network analysis  Behavior self-organization  Exceptions and outlier mining

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Behavioral intelligence emergence behavioral occurrences, evolution and life cycles impact of particular behavioral rules and patterns on behavioral evolution and intelligence emergence define and model behavioral rules, protocols and relationships, and their impact on behavioral evolution and intelligence emergence

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Behavioral network intrinsic mechanisms inside a network behavioral rules, interaction protocols, convergence and divergence of associated behavioral itemsets effects such as network topological structures, linkage relationships, and impact dynamics Community formation, pattern, dynamics and evolution

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Behavioral simulation observe the dynamics, the impact of rules/protocols/patterns, behavioral intelligence emergence, and the formation and dynamics of social behavioral network

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM  Large-scale behavior network  Behavior convergence and divergence  Behavior learning and adaptation  Group behavior formation and evolution  Behavior interaction and linkage  Artificial behavior system  Computational behavior system  Multi-agent simulation

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Behavioral presentation presentation means and tools describe the motivation and the interest of stakeholders on the particular behavioral data Traditional behavior pattern presentation visual behavioral presentation

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM  Rule-based behavior presentation  Flow visualization  Sequence visualization  Parallel visualization  Dynamic group formation  Dynamic behavior impact evolution  Visual behavior network  Behavior lifecycle visualization  Temporal-spatial relationship  Dynamic factor tuning, configuration and effect analysis  Behavior pattern emergence visualization  Distributed, linkage and collaborative visualization

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM BIA general process

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Theoretical Underpinnings Methodological support, Fundamental technologies, and Supporting techniques and tools

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Applications Trading Behavior Analysis Customer-Officer Interaction Analysis in Social Security Areas Facial behavior analysis Online user behavior analysis …

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Trading Behavior Analysis Cao L., Ou, Y. Market microstructure patterns powering trading and surveillance agents. Journal of Universal Computer Sciences, 14(14): , (1) indicating the direction, probability and size of an order to be traded, (2) reflecting an order’s dynamics during its lifecycle

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Customer-Officer Interaction Analysis in Social Security Areas  Cao, L., Zhao, Y., Zhang, C. (2008), Mining Impact-Targeted Activity Patterns in Imbalanced Data, IEEE Trans. Knowledge and Data Engineering, IEEE,, Vol. 20, No. 8, pp , 2008.

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Facial behavior analysis Pohsiang Tsai; Tom Hintz, Tony Jan, Longbing Cao. A New Multimodal Biometrics for Personal Identification, Pattern Recognition Letters (to appear)

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM References Cao L. From Behavior to Solutions: the Behavior Informatics and Analytics Approach, Information Sciences, to appear. Cao, L., Zhao, Y., Zhang, C. Mining impact-targeted activity patterns in imbalanced data, IEEE Trans. on Knowledge and Data Engineering, Vol. 20, No. 8, pp , 2008 Cao, L., Zhao, Y., Zhang, C., Zhang, H. Activity mining: from activities to actions, International Journal of Information Technology & Decision Making, 7(2), pp , 2008 Cao L., Ou, Y. Market microstructure patterns powering trading and surveillance agents. Journal of Universal Computer Sciences, 2008.

The Smart Lab: datamining.it.uts.edu.au BIA: BIA: Behavior Informatics and Analytics 15 December 2008Cao, L: BIA at DDDM2008 Joint with ICDM Thank you! Longbing CAO Faculty of Engineering and IT University of Technology, Sydney, Australia Tel: Fax: Homepage: www-staff.it.uts.edu.au/~lbcao/www-staff.it.uts.edu.au/~lbcao/ The Smart Lab: datamining.it.uts.edu.audatamining.it.uts.edu.au