Coupled Behavior Analysis with Applications

Transcription

1 Coupled Behavior Analysis with Applications Professor Longbing Cao ( 操 龙 兵 ) Director, Advanced Analytics Institute University of Technology Sydney, Australia www-staff.it.uts.edu.au/~lbcao

2 Agenda Why coupled behaviors? What is behavior? What are coupled behaviors? What is coupled behavior analysis (CBA)? Combined mining for high-impact behavior analysis Coupled HiddenMarkov Model-based abnormal behavior analysis Open issues and prospects

3 Why Coupled Behaviors?

4 An example Why does this stock go so crazily?

5 Short-term manipulation behaviors as cause Behavior exterior presentation Possible behavior interior driver

6 Behaviors of associated accounts as the driver of the price movement Group behaviors Group behavior interior-driven price movement

7 What makes multiple behaviors different? Key factors: Multiple actors Multiple behaviors Multiple properties Coupling relationships Organizational factors

8 How are CB handled by existing techniques? Behavior exterior analysis Time series analysis Multiple time series analysis Behavior interior analysis Frequent pattern mining Sequence analysis Coupled sequence analysis

9 Coupledbehaviors are ubiquitous Relevant projects in UTS Advanced Analytics Institute Insurance business analytics Public service business analytics Education student analytics Investment business analytics Banking business analytics Financial business analytics

10 What is Behavior? Longbing Cao, In-depth Behavior Understanding and Use: the Behavior Informatics Approach, Information Science, 180(17); ,

11 Abstract Behavior Model

12 An abstract behavior model Demographics and circumstancesof 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. Impact, costs, risk and trust of behavior/behavior network

13 Behavior Visual Descriptor Intra-relationships within Behavior

14 Vector-oriented behavior pattern analysis 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

15 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

16 Behavior informatics Concept Map Behavior Representation & Reasoning Behavior Learning & Mining

17 Behavior learning/mining process

18 What is Coupled Behavior? Longbing Cao, In-depth Behavior Understanding and Use: the Behavior Informatics Approach, Information Science, 180(17); ,

19 Relationship crossing behaviors

20 Coupling relationships From temporal aspect From inferential aspect From combinational aspect From party aspect

21 Basic Behavior Patterns Tracing: Different actions with sequential order. { a a a },,, n 1 2 Consequence: Different actions have causalities in occurrence. { ai a j } Synchronization: Different actions occur at the same time. {,, } a1 a n Combination: Different actions occur in concurrency. { a1 a,, 2 an}

22 Exclusion: Different actions occur mutually exclusively. { a a a } 1 2,, { ai a j} n Precedence: Different actions have required precedence And more to be explored Sequential Combination Parallel Combination Nested Combination Fuzzy or probabilistic Combination A B C A B C

23 What is the Coupled Behavior Analysis (CBA) problem? LongbingCao, YumingOu, Philip S Yu. Coupled Behavior Analysis with Application, IEEE Trans. Knowledge and Data Engineering. LongbingCao, In-depth Behavior Understanding and Use: the Behavior Informatics Approach, Information Science, 180(17); ,

24 Customer behaviors Customer a i snbehaviors B i : {b i1, b i2,,b in } M customers behaviors B 1 : {b 11, b 12,,b 1n } B 2 : {b 21, b 22,,b 2n } B m : {b m1, b m2,,b mn }

25 Behavior Feature Matrix

26 Behavior Intra-relationship

27 Behavior Inter-relationship

28 Behavior Relationship

29 Behavior Behavior Analysis

30 An Example of Stock Market Transactional Data Behavior Feature Matrix B1 B2 B3 B4 B5 B6 B7 B8

31 Existing approaches M customers behaviors B 1 : {sell, buy} B 2 : {buy, sell, sell} B 3 : {buy} B 4 : {buy} B 5 : {buy}

32 Complex behavior pattern analysis (sell,sell_price,volume_small,long_interval,non-frequent) % (sell,sell_price,volume_small,long_interval,non-frequent) (action_other,price_other,volume_other,interval_other,nonfrequent) % (buy_withdraw,price_other,withdraw_part,short_withdraw interval,non-frequent) % (action_other,price_other,volume_other,interval_other,non-frequent) % (action_other,price_other,volume_other,interval_other,non-frequent) (action_other,price_other,volume_other,interval_other,non-frequent) % (buy_withdraw,price_other,withdraw_part,long_withdraw interval,non-frequent) % (buy_withdraw,price_other,withdraw_part,long_withdrawinterval,non-frequent) (buy_withdraw,price_other,withdraw_part,long_withdraw interval,non-frequent) % (buy_withdraw,price_other,withdraw_part,long_withdrawinterval,non-frequent) (action_other,price_other,volume_other,interval_other,non-frequent) % (buy_withdraw,price_other,withdraw_part,long_withdrawinterval,non-frequent) (action_other,price_other,volume_other,interval_other,non-frequent) (action_other,price_other,volume_other,interval_other,non-frequent) % (sell_withdraw,price_other,withdraw_part,short_withdraw interval,non-frequent) % (buy,buy_price_lastor buy_price_limitor buy_price_sell,volume_small,long_interval,non-frequent) % (buy,buy_price_lastor buy_price_limitor buy_price_sell,volume_small,long_interval,non-frequent) (action_other,price_other,volume_other,interval_other,non-frequent) % (sell_withdraw,price_other,withdraw_part,long_withdraw interval,non-frequent) % (sell_withdraw,price_other,withdraw_part,long_withdrawinterval,non-frequent) (sell,sell_price,volume_small,long_interval,non-frequent) % (sell_withdraw,price_other,withdraw_part,long_withdrawinterval,non-frequent) (sell_withdraw,price_other,withdraw_part,long_withdraw interval,non-frequent) % (sell_withdraw,price_other,withdraw_part,long_withdrawinterval,non-frequent) (action_other,price_other,volume_other,interval_other,non-frequent) % (sell_withdraw,price_other,withdraw_part,long_withdrawinterval,non-frequent) (action_other,price_other,volume_other,interval_other,non-frequent) (action_other,price_other,volume_other,interval_other,non-frequent) %

33 Hierarchical clustering analysis

34 Combined Pattern Mining for High Impact Behavior Analysis Longbing Cao, Huaifeng Zhang, Yanchang Zhao, Dan Luo, Chengqi Zhang. Combined Mining: Discovering Informative Knowledge in Complex Data, accepted by IEEE Trans. SMC Part B LongbingCao. Zhao Y., Zhang, C. Mining Impact-Targeted Activity Patterns in Imbalanced Data, IEEE Trans. on Knowledge and Data Engineering, 20(8): , 2008.

35 Combined Pattern Pairs A combined rule pair is composed of two contrasting rules. For customers with same characteristics U, different policies/campaigns, V 1 and V 2, can result in different outcomes, T 1 and T 2.

36 Interestingness of Pattern Pairs

37 Combined Pattern Clusters Based on a combined rule pair, related combined rules can be organized into a cluster to supplement more information to the rule pair. The rules in cluster C have the same U but different V, which makes them associated with various results T.

38 Interestingness of Rule Pair/Cluster dist(): the dissimilarity between the descendants of R 1 and R 2 The interestingness of combined rule pair/cluster is decided by both the interestingness of rules and the most contrasting rules within the pair/cluster. A cluster made of contrasting confident rules is interesting, because it explains why different results occur and what can be done to produce an expected result or avoid an undesirable consequence.

39 Extended Combined Pattern Pairs

40 Conditional P-S ratio

41 Extended Combined Pattern Clusters

42 Impact

43 Combined Demographics + Behavior Analysis Longbing Cao, Huaifeng Zhang, Yanchang Zhao, Dan Luo, Chengqi Zhang. Combined Mining: Discovering Informative Knowledge in Complex Data, IEEE Trans. SMC Part B. Longbing Cao. Zhao Y., Zhang, C. Mining Impact-Targeted Activity Patterns in Imbalanced Data, IEEE Trans. on Knowledge and Data Engineering, 20(8): , Yanchang Zhao, Huaifeng Zhang, Longbing Cao Chengqi Zhang. Combined Pattern Mining: from Learned Rules to Actionable Knowledge, Australian AI2008.

44 Combined Pattern Mining Type A: Demographics differentiated combined pattern Customerswith the same actionsbut different demographics different classes/business impact

45 Combined Pattern Mining Type B: Actiondifferentiatedcombined pattern Customerswith the same demographicsbut taking different actions different classes/business impact

46 There were 7,711 association rules before removing redundancy of combined rules. After removing redundancy of combined rules, 2,601 rules were left, which built up 734 combined rule clusters. After removing redundancy of combined rule clusters, 98 rule clusters with 235 rules remained, which was within the capability of human beings to read.

47

48 Behavior 1 Behavior 2 Demographic 1 Low value High value Demographic 2 High value Low value

49 An Example of Extended Combined Pattern Cluster

50 Identifying high impact behavior in behavior evolution

51 Coupled Hidden Markov Model- based Abnormal Coupled Behavior Analysis LongbingCao, YumingOu, Philip S Yu. Coupled Behavior Analysis with Application, IEEE Trans. Knowledge and Data Engineering. Cao, L., OuY, Yu PS, Wei G. Detecting Abnormal Coupled Sequences and Sequence Changes in Group-based Manipulative Trading Behaviors, KDD2010.

52 Pool manipulation

53 Coupled behavior analysis

54 Construct behaviorsequences Data structure 1:

55 Data structure 2:

56 CHMM Based Coupled Sequence Modeling Coupled behavior sequences Multiple sequences Coupling relationship Behavior properties

57 CBA -CHMM

58 Framework: abnormal CBA

59

60

61 CHMM model:

62 Adaptive CHMM for Detecting Sequence Changes

63 The Algorithms

64

65 Benchmark Models HMM-B: Buy-based HMM HMM-S: Sell-based HMM HMM-T: Trade-based HMM IHMM: HMM-B + HMM-S + HMM-T CHMM: CHMM(buy, sell, trade) ACHMM: Adaptive CHMM(buy, sell, trade)

66 Evaluation Technical performance Business performance

67

68 Business Performance

69 Computational cost N-heads dynamic programming

70 Prospects

71 Sequence analysis Frequent Pattern mining Impact-oriented: - Positive - Negative - Multi-level level - Mixed - Evolution Coupled behavior analysis Group Behavior Pattern mining Event detection Community discovery

72 Novel Behavior Pattern Mining Semi-supervised coupled behavior analysis 1: Coupling relationship analysis 2: coupling-oriented Pattern mining

73 Behavior Informatics-SIG: 73 Cao, L: BI at DDDM2008 Joint with ICDM2008

74 References LongbingCao, YumingOu, Philip S Yu. Coupled Behavior Analysis with Applications, accepted by IEEE Trans. on Knowledge and Data Engineering. LongbingCao, HuaifengZhang, YanchangZhao, Dan Luo, ChengqiZhang. Combined Mining: Discovering Informative Knowledge in Complex Data, accepted by IEEE Trans. SMC Part B LongbingCao, YumingOu, Philip S YU, Gang Wei. Detecting Abnormal Coupled Sequences and Sequence Changes in Group-based Manipulative Trading Behaviors, KDD2010, LongbingCao, In-depth Behavior Understanding and Use: the Behavior Informatics Approach, Information Science, 180(17); , 2010 LongbingCao, YanchangZhao, ChengqiZhang. Mining Impact-Targeted Activity Patterns in Imbalanced Data, IEEE Trans. on Knowledge and Data Engineering, 20(8): , 2008

75 Thank you for your attention Longbing Cao www-staff.it.uts.edu.au/~lbcao