WO2016004744A1 - 基于复杂对应系统的用户行为一致性度测量方法 - Google Patents

基于复杂对应系统的用户行为一致性度测量方法 Download PDF

Info

Publication number
WO2016004744A1
WO2016004744A1 PCT/CN2014/095859 CN2014095859W WO2016004744A1 WO 2016004744 A1 WO2016004744 A1 WO 2016004744A1 CN 2014095859 W CN2014095859 W CN 2014095859W WO 2016004744 A1 WO2016004744 A1 WO 2016004744A1
Authority
WO
WIPO (PCT)
Prior art keywords
behavior
matrix
relationship
user
consistency
Prior art date
Application number
PCT/CN2014/095859
Other languages
English (en)
French (fr)
Inventor
蒋昌俊
陈闳中
闫春钢
丁志军
王咪咪
赵培海
Original Assignee
同济大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 同济大学 filed Critical 同济大学
Priority to US15/325,184 priority Critical patent/US20170169350A1/en
Priority to DE112014006799.7T priority patent/DE112014006799T5/de
Publication of WO2016004744A1 publication Critical patent/WO2016004744A1/zh
Priority to AU2017100012A priority patent/AU2017100012A4/en

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063Operations research, analysis or management
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/28Databases characterised by their database models, e.g. relational or object models
    • G06F16/284Relational databases
    • G06F16/285Clustering or classification
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/28Databases characterised by their database models, e.g. relational or object models
    • G06F16/284Relational databases
    • G06F16/288Entity relationship models
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/16Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/29Graphical models, e.g. Bayesian networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/04Inference or reasoning models
    • G06N5/048Fuzzy inferencing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00Computing arrangements based on specific mathematical models
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising

Definitions

  • the invention relates to user behavior consistency measurement and can be applied to internet payment platform security.
  • the object of the present invention is to overcome the deficiencies of the prior art, and to measure the behavior consistency between the user behavior model and the expected model, perform specific classification analysis on the complex corresponding behavior relationship, determine the behavior corresponding features of each complex class, and solve the problem corresponding to the crossover repetition.
  • the behavioral consistency problem using the relevant knowledge of the matrix, calculates the behavioral consistency of the model, and measures the behavioral consistency compliance with complex correspondence.
  • a method for measuring user behavior consistency based on a complex corresponding system characterized in that the whole scheme is divided into three phases:
  • Step 1-1 on the basis of the existing workflow network, subdivide the cross-order relationship and refine the behavioral contour relationship;
  • Step 1-2 analyzing complex correspondences, classifying complex correspondences, and determining behavior characteristics of each class;
  • Step 1-3 at the same time, analyzing the transfer dependencies between user activities according to the indirect relationship between users;
  • Step 2-1 determining the correlation between the five types of correspondences according to the classification of the complex correspondences completed in step 1-2 and the behavior characteristics of each class;
  • Step 2-2 establishing a behavior profile relationship of the user extension according to the behavior profile relationship refined in step 1-1;
  • Step 2-4 based on steps 2-2 and 2-3, construct a user behavior relationship matrix diagram
  • Step 3-1 according to the five types of user complex correspondence classes determined in step 2-1 and the behavior relationship matrix diagram established in step 2-4, the user behavior relationship matrix is decomposed;
  • Step 3-2 calculating the behavior consistency between the user model and the expected model according to the correspondence between the actual model of the user and the expected model.
  • the consistent behavior relationship shows the consistent part of the user activity, and the whole behavioral relationship is characterized by the area of the behavior matrix.
  • Figure 3 is the behavior diagram of Figure 2
  • Figure 4 is an exploded view of Figure 3
  • Figure 5 is a flow chart of Algorithm 1
  • Figure 6 is an algorithm 2 flow chart
  • the specific implementation method analyzes the internal behavior relationship of the user in detail, establishes the outline of the user behavior relationship, and distinguishes and classifies the complex correspondence relationship, and gives the user line based on the complex correspondence relationship.
  • the framework can effectively distinguish complex correspondences and make more accurate decisions on behavioral correspondences. Effectively distinguish and calculate complex correspondences, solve the behavior consistency measurement problem of complex corresponding model pairs, and greatly shorten the computing time.
  • the system structure diagram of the user behavior consistency measurement method is shown in Figure 1.
  • the whole scheme is divided into three phases: the first phase analyzes the complex correspondence characteristics according to the existing user behavior model, the second phase establishes the behavior contour according to the user behavior characteristics, constructs the user behavior relationship matrix, and the third stage is completed according to the user complex correspondence characteristics.
  • the user behavior matrix is decomposed, the user behavior consistency is calculated, and the degree of consistency between the user behavior and the expected behavior is detected.
  • Step 1-1 on the basis of the existing workflow network, subdivide the cross-order relationship and refine the behavior contour relationship.
  • Step 1-2 Analyze complex correspondences, classify complex correspondences, and determine behavior characteristics of each class.
  • Steps 1-3 while analyzing the transfer dependencies between user activities based on the indirect relationship between users.
  • steps 1-1, 1-2 and 1-3 above are performed in parallel.
  • step 2-1 according to the classification of the complex correspondences completed in step 1-2 and the behavior characteristics of each class, the correlation between the five types of correspondences is determined.
  • step 2-2 the behavior profile relationship of the user extension is established according to the behavior profile relationship refined in step 1-1.
  • Step 2-4 based on steps 2-2 and 2-3, construct a user behavior relationship matrix diagram.
  • Step 3-1 according to the five types of user complex correspondence classes determined in step 2-1 and the behavior relationship matrix diagram established in step 2-4, the user behavior relationship matrix is decomposed (see Algorithm 1 for details).
  • Step 3-2 Calculate the behavior consistency between the user model and the expected model according to the correspondence between the actual model of the user and the expected model (see Algorithm 2 for details).
  • the consistent behavioral relationship shows the consistent part of the user's activity.
  • step (4) MD A0 in accordance with a second MD B0 corresponding set ⁇ a m + 1, a m + 2, ... a l ⁇ , taking the first MD A0 1 ⁇ (m) rows and (m + 1) ⁇ (l)
  • the m ⁇ (lm) order matrix composed of columns and its transposed matrix are denoted as module 2, and step (4) is performed.
  • step (5) the previous step modeled until MD A0 in the p-th and MD B0 corresponding to the set ⁇ a s + 1, ... a n ⁇ , taking MD A0 first 1 ⁇ (m) rows and (s + 1) ⁇
  • the m ⁇ (ns) order matrix composed of (n) columns and its transposed matrix are denoted as module p, and step (5) is performed.
  • MD A0 in accordance with a second set of MD B0 corresponding to ⁇ a m + 1, a m + 2, ... a l ⁇ , taking the first MD A0 (m + 1) ⁇ (l ) rows and (m
  • the (lm) order matrix composed of +1) ⁇ (l) columns is recorded as module p+1, and step (6) is performed.
  • step (4) the (lm) ⁇ (ns) order matrix composed of (m+1) ⁇ (l) rows and (s+1) ⁇ (n) columns in MD A0 and its transposed matrix Record as module p+2 and perform step (7).
  • step (9) is performed.
  • the behavior relationship matrix diagrams MD a , MD b , MD c , MD d (shown in Figure 3) of Figures 2(a), (b), (c), and (d) are respectively obtained, and then according to the algorithm.
  • Steps (1)-(9) of 2 are respectively decomposed, taking MD a and MD b as an example, as shown in FIG. 4 .
  • the step (10) of the algorithm 2 the degree of consistency of (a) and (b) in Fig.
  • the behavior of the user (a) shown in FIG. 2 is 75% consistent with the behavior of (b), and the behavior of the user (b) shown in FIG. 2 is 80% consistent with the behavior of (c), as shown in FIG.
  • the behavior of user (c) is approximately 81% consistent with the behavior of (d), which is high, indicating that the user behavior is consistent with the expected behavior, and we determine that the user behavior is legal.

Abstract

一种基于复杂对应系统的用户行为一致性度测量方法,应用于互联网支付平台安全。整个方案分为三个阶段:第一阶段根据现有用户行为模型分析复杂对应关系特征,第二阶段根据用户行为特征建立行为轮廓,构建用户行为关系矩阵,第三阶段根据用户复杂对应特征,完成用户行为矩阵分解,计算用户行为一致性度,检测用户行为与预期行为的一致程度。对用户的内部行为关系进行了较细致的分析,建立了用户行为关系的轮廓,并对复杂对应关系进行区分和分类,给出了基于复杂对应关系用户行为的一致性测量和分析构架。有效地将复杂对应关系进行区分和计算,解决了存在复杂对应模型对的行为一致性测度问题,并大大缩短了运算时间。

Description

基于复杂对应系统的用户行为一致性度测量方法 技术领域
本发明涉及用户行为一致性度测量,可应用于互联网支付平台安全。
背景技术
随着计算机的飞速发展,网上支付平台的应用越来越广泛,对于用户在支付过程中的行为一致性检测技术的要求也越来越严格。
由于系统设计师和建模者对相同的真实世界现象所持有的观点的不同,从而导致不同模型的建立。模型的一致性关系到模型元素在模型匹配情况下的匹配的语义学。那么存在复杂对应的情况便不言而喻,统计表明对于流程模型中存在的对应中,有超过40℅是复杂对应的,超过7℅是有着交叉重复对应的。如何对电子交易过程中用户的行为和预期行为进行一致性分析,在存在于复杂系统中的模型间便显得至关重要。
先前就两个模型(即测量用户行为模型、预期模型)之间的一致性有过一些研究,提出了如迹匹配、互模拟、行为轮廓等测量方法(见后面的批注[1-5]),但是这些方法在复杂对应方面,未能有效区分行为间的复杂对应的情况,从而在计算精度上面大打折扣。
提供以下索引,索引所对应的公开文献为与本发明技术方案接近或者相关技术,并也视为本发明说明书的组成部分。因此,本发明技术方案中涉及的技术术语以及技术方案实施所依赖的在先技术可参见如下资料:
[1]Matthias Weidlich,Jan Mendling,Mathias Weske.Efficient consistency measurement based on behavioral pro fi les of process models[J].IEEE Transactions on Software Engineering,2011,37(3):410–429.
[2]Matthias Weidlic,Behavioural profiles---a relational approach to behaviour consistency[DB/OL].Institutional Repository of the University of Potsdam:URL http://opus.kobv.de/ubp/volltexte/2011/5559/URN urn:nbn:de:kobv:517‐opus‐55590,2011.
[3]Sergey Smirnov,Matthias Weidlich,Jan Mendling.Business Process Model Abstraction Based on Behavioral Profiles[C].Heidelberg:Springer  Verlag,2010:1-16.
[4]Matthias Weidlich,Mathias Weske,Jan Mendling.Change Propagation in Process Models Using Behavioural Pro fi les[C].Washington:IEEE Computer Society Washington,2009:33-40.
[5]Matthias Weidlich,Jan Mendling.Perceived consistency between process models[J].Information Systems,2012,37(2):80-98.
[6]吴哲辉,Petri网导论[M].机械工业出版社,2006年。
发明内容
本发明目的在于克服现有技术的不足,用于测量用户行为模型与预期模型的行为一致性,对复杂对应的行为关系进行具体分类分析,确定各个复杂类的行为对应特征;解决含有交叉重复对应的行为一致性问题,利用矩阵的相关知识,计算了模型的行为一致性,测量了含有复杂对应关系的行为一致性服从度。
为此,给出的技术方案为:
一种基于复杂对应系统的用户行为一致性度测量方法,其特征在于,整个方案分为三个阶段:
第一阶段具体实施步骤:
步骤1-1,在已有工作流网的基础上,对交叉序关系进行细分,细化行为轮廓关系;
步骤1-2,分析复杂对应关系,将复杂对应关系分类,确定各个类的行为特征;
步骤1-3,同时根据用户之间的间接关系分析用户活动间的传递依赖关系;
以上步骤1-1、1-2和1-3是并列进行着;
第二阶段具体实施步骤:
步骤2-1,根据步骤1-2完成的复杂对应关系的分类及其每个类的行为特征,确定五类对应关系之间的相关性;
步骤2-2,根据步骤1-1细化的行为轮廓关系,建立用户扩展的行为轮廓关系;
步骤2-3,在步骤2-2的基础上并结合步骤1-3,根据公式
Figure PCTCN2014095859-appb-000001
(i,j=1,2,…,n)把用户行为关系转化成矩阵元素(其中aij为行为关系矩阵中的元素);
步骤2-4,在步骤2-2和2-3的基础上,构建用户行为关系矩阵图;
其构造步骤如下所示(从矩阵MD1→MD2→MD3→MD4…→MDn→MD):
Figure PCTCN2014095859-appb-000002
Figure PCTCN2014095859-appb-000003
Figure PCTCN2014095859-appb-000004
Figure PCTCN2014095859-appb-000005
……
Figure PCTCN2014095859-appb-000006
Figure PCTCN2014095859-appb-000007
第三阶段具体实施步骤:
步骤3-1,根据步骤2-1确定的五类用户复杂对应类以及步骤2-4建立的行为关系矩阵图,将用户行为关系矩阵进行分解;
步骤3-2,根据用户实际模型与预期模型的对应关系,计算用户模型与预期模型的行为一致性,
Figure PCTCN2014095859-appb-000008
其中,一致的行为关系表现出用户活动的一致的部分,用行为矩阵的面积来刻画其整个一致的行为关系,一致度值越高代表该用户行为与预期行为越一致,一致度值越低代表该用户行为与预期行为越不一致,当一致度特别低时,怀疑该用户行为为非法行为。
附图说明
图1系统架构图
图2业务流程Petri网图
图3是图2的行为关系图
图4是图3的分解图
图5是算法1流程图
图6是算法2流程图
具体实施方式(案例)对用户的内部行为关系进行了较细致的分析,建立了用户行为关系的轮廓,并对复杂对应关系进行区分和分类,给出了基于复杂对应关系用户行 为的一致性测量和分析构架,如图1所示。该构架能够有效区分复杂对应关系,并依此对行为对应关系做到更为精确的判定。有效地将复杂对应关系进行区分和计算,解决了存在复杂对应模型对的行为一致性测度问题,并大大缩短了运算时间。
用户行为一致性度测量方法系统结构图,如图1所示。整个方案分为三个阶段:第一阶段根据现有用户行为模型分析复杂对应关系特征,第二阶段根据用户行为特征建立行为轮廓,构建用户行为关系矩阵,第三阶段根据用户复杂对应特征,完成用户行为矩阵分解,计算用户行为一致性度,检测用户行为与预期行为的一致程度。
第一阶段具体实施步骤:
步骤1-1,在已有工作流网的基础上,对交叉序关系进行细分,细化行为轮廓关系。
步骤1-2,分析复杂对应关系,将复杂对应关系分类,确定各个类的行为特征。
步骤1-3,同时根据用户之间的间接关系分析用户活动间的传递依赖关系。
其中:以上步骤1-1、1-2和1-3是并列进行着的。
第二阶段具体实施步骤:
步骤2-1,根据步骤1-2完成的复杂对应关系的分类及其每个类的行为特征,确定五类对应关系之间的相关性。
步骤2-2,根据步骤1-1细化的行为轮廓关系,建立用户扩展的行为轮廓关系。
步骤2-3,在步骤2-2的基础上并结合步骤1-3,根据公式
Figure PCTCN2014095859-appb-000009
(i,j=1,2,…,n)把用户行为关系转化成矩阵元素(其中aij为行为关系矩阵中的元素)。
步骤2-4,在步骤2-2和2-3的基础上,构建用户行为关系矩阵图。
其构造步骤如下所示(从矩阵MD1→MD2→MD3→MD4…→MDn→MD):
Figure PCTCN2014095859-appb-000010
Figure PCTCN2014095859-appb-000011
Figure PCTCN2014095859-appb-000012
Figure PCTCN2014095859-appb-000013
……
Figure PCTCN2014095859-appb-000014
Figure PCTCN2014095859-appb-000015
第三阶段具体实施步骤:
步骤3-1,根据步骤2-1确定的五类用户复杂对应类以及步骤2-4建立的行为关系矩阵图,将用户行为关系矩阵进行分解(具体见算法1)。
步骤3-2,根据用户实际模型与预期模型的对应关系,计算用户模型与预期模型的行为一致性(具体见算法2)。
Figure PCTCN2014095859-appb-000016
其中,一致的行为关系表现出用户活动的一致的部分,我们用行为矩阵的面积来刻画其整个一致的行为关系,一致度值越高代表该用户行为与预期行为越一致,一致度值越低代表该用户行为与预期行为越不一致,当一致度特别低时,我们怀疑该用户行为为非法行为。
算法1行为关系矩阵图中元素的求解算法。(具体流程见图5)
输入:两个工作流网N1=(P1,T1;F1)和N2=(P2,T2;F2),其中他们中有着对应关系的 变迁集A={a1,a2,…,an}、B={b1,b2,…,bm}、
Figure PCTCN2014095859-appb-000017
Figure PCTCN2014095859-appb-000018
进行排序的行为关系矩阵MDA0和MDB0
输出:行为关系矩阵图MDA、MDB中的元素aij(i,j=1,2,…,n)、bij(i,j=1,2,…,m)。
(1)先确定MDA中对角线的元素aii(i=1,2,…,n),依次判断ai(i=1,2,…,n)是否处于环结构中,若ai不在环结构中,那么输出aii=2,执行步骤(2);否则输出aii=0,执行步骤(2)。
(2)再确定ai,i+1、ai+1,i(i=1,2,…,n-1)的值。在网N1中,依次计算ai与ai+1的行为关系,然后将行为关系转化为整数p,输出ai,i+1=ai+1,i=p,执行步骤(3)。
(3)再确定ai,i+2、ai+2,i(i=1,2,…,n-2)的值。若ai,i+1≠ai+1,i+2,输出ai,i+2=ai+2,i=min{ai,i+1,ai+1,i+2};否则,若ai,i+1=ai+1,i+2=1,那么输出ai,i+2=ai+2,i=1;否则,若ai,i+1=ai+1,i+2≠1,那么判断ai与ai+2的行为关系,并转化为行为关系数值q,输出ai,i+2=ai+2,i=q,执行步骤(4)。
(4)同理,确定ai,i+h、ai+h,i(i=1,2,…,n-h)(h=3…,n-1),输出ai,i+h=ai+h,i,直到最后一个元素a1n,算法终止。
同理我们根据算法1,可以计算出MDB中的元素bij(i,j=1,2,…,m),从而得到矩阵MDB
算法2一致性度的求解算法。(具体流程见图6)
输入:两个工作流网N1=(P1,T1;F1)和N2=(P2,T2;F2),他们的行为关系矩阵MDA0和MDB0(由算法1求得)。
输出:一致性度BP。
(1)先根据MDA0与MDB0中变迁集的对应关系,将MDA0和MDB0分别分为p和q个 对应集合,将MDA0依次标记为{a1,a2,…am}、{am+1,am+2,…al}…{as+1,…an},执行步骤(2)
(2)先根据MDA0中第一个与MDB0对应的集合{a1,a2,…am},取MDA0中前m阶方阵记为模块1,并执行步骤(3)。
(3)根据MDA0中第二个与MDB0对应的集合{am+1,am+2,…al},取MDA0中第1→(m)行和(m+1)→(l)列组成的m×(l-m)阶矩阵及其转置矩阵记为模块2,并执行步骤(4)。
(4)仿照之前的步骤,直到MDA0中第p个与MDB0对应的集合{as+1,…an},取MDA0中第1→(m)行和(s+1)→(n)列组成的m×(n-s)阶矩阵及其转置矩阵记为模块p,并执行步骤(5)。
(5)根据MDA0中第二个与MDB0对应的集合{am+1,am+2,…al},取MDA0中第(m+1)→(l)行和(m+1)→(l)列组成的(l-m)阶方阵记为模块p+1,并执行步骤(6)。
(6)仿照步骤(4),将MDA0中(m+1)→(l)行和(s+1)→(n)列组成的(l-m)×(n-s)阶矩阵及其转置矩阵记为模块p+2,并执行步骤(7)。
(7)一直这样进行下去,直到MDA0中第p个与MDB0对应的集合{as+1,…an},取s+1→n行和s+1→n列组成的(n-s)阶方阵记为模块
Figure PCTCN2014095859-appb-000019
并执行步骤(8)。
(8)若p=q,同理将MDB0也分解为
Figure PCTCN2014095859-appb-000020
对应模块,并标记模块名从
Figure PCTCN2014095859-appb-000021
Figure PCTCN2014095859-appb-000022
则执行步骤(10);否则若p≠q,将MDB0中非重复对应关系也分解为
Figure PCTCN2014095859-appb-000023
对应模块,执行步骤(9)。
(9)锁定重复对应的变迁集合,以重复对应的集合组成的区域依次记为模块
Figure PCTCN2014095859-appb-000024
执行步骤(10)。
(10)在MDA0中,依次对模块
Figure PCTCN2014095859-appb-000025
中的矩阵元素进行排查,找出其ai、ai与MDB0相同模块中不同的元素bi、bj,若p=q,输出一致性度BP,算法终止;否则若p≠q,则锁定模块1c、2c…,(q-p)c,输出一致性度BP,算法终止。
下面给出图2的一个例子。
根据算法1,分别得到图2(a)、(b)、(c)、(d)的行为关系矩阵图MDa、MDb、MDc、MDd(如图3所示),然后根据算法2的步骤(1)-(9),分别对其进行分解,以MDa、MDb为例,如图4所示。根据算法2的步骤(10),可得图2中(a)和(b)的一致性度为:
Figure PCTCN2014095859-appb-000026
同理可得到图2中(b)和(c)的一致性度为:
Figure PCTCN2014095859-appb-000027
而图中(c)和(d)中,有A~{A1,AB1,AB2}且有B~{AB1,AB2},有图2中(c)和(d)的轮廓一致性度为:
Figure PCTCN2014095859-appb-000028
图2所示的用户(a)的行为与(b)的行为一致度达到75%,图2所示的用户(b)的行为与(c)的行为一致度达到80%,图2所示的用户(c)的行为与(d)的行为一致度约达到81%,都比较高,表明该用户行为与预期行为一致,我们判定该用户行为为合法行为。
本发明的创新点
1.利用行为轮廓技术,将用户行为模式一致性量化。
2.对用户复杂行为关系进行分类,并确定了各个复杂类的行为特征及性质。
3.提出行为矩阵的方法,将模型对间的行为关系转化为行为关系矩阵的元素,缩短了计算时间。
4.区分了交叉重复对应的情况,提高精确度,解决了存在交叉重复模型对的行为一致性测度问题。

Claims (3)

  1. 一种基于复杂对应系统的用户行为一致性度测量方法,其特征在于,整个方案分为三个阶段:
    第一阶段具体实施步骤:
    步骤1-1,在已有工作流网的基础上,对交叉序关系进行细分,细化行为轮廓关系;
    步骤1-2,分析复杂对应关系,将复杂对应关系分类,确定各个类的行为特征;
    步骤1-3,同时根据用户之间的间接关系分析用户活动间的传递依赖关系;
    以上步骤1-1、1-2和1-3是并列进行着;
    第二阶段具体实施步骤:
    步骤2-1,根据步骤1-2完成的复杂对应关系的分类及其每个类的行为特征,确定五类对应关系之间的相关性;
    步骤2-2,根据步骤1-1细化的行为轮廓关系,建立用户扩展的行为轮廓关系;
    步骤2-3,在步骤2-2的基础上并结合步骤1-3,根据公式
    Figure PCTCN2014095859-appb-100001
    (i,j=1,2,…,n)把用户行为关系转化成矩阵元素(其中aij为行为关系矩阵中的元素);
    步骤2-4,在步骤2-2和2-3的基础上,构建用户行为关系矩阵图;
    其构造步骤如下所示(从矩阵MD1→MD2→MD3→MD4…→MDn→MD):
    Figure PCTCN2014095859-appb-100002
    Figure PCTCN2014095859-appb-100003
    第三阶段具体实施步骤:
    步骤3-1,根据步骤2-1确定的五类用户复杂对应类以及步骤2-4建立的行为关系矩阵图,将用户行为关系矩阵进行分解;
    步骤3-2,根据用户实际模型与预期模型的对应关系,计算用户模型与预期模型的行为一致性,
    计算公式:
    Figure PCTCN2014095859-appb-100004
    其中,一致的行为关系表现出用户活动的一致的部分,用行为矩阵的面积来刻画其整个一致的行为关系,一致度值越高代表该用户行为与预期行为越一致,一致度值越低代表该用户行为与预期行为越不一致,当一致度特别低时,怀疑该用户行为为非法行为。
  2. 如权利要求1所述的基于复杂对应系统的用户行为一致性度测量方法,其特征在于,步骤3-1中,所述将用户行为关系矩阵进行分解,其行为关系矩阵图中元素的求解算法为:
    输入:两个工作流网N1=(P1,T1;F1)和N2=(P2,T2;F2),其中他们中有着对应关系的 变迁集A={a1,a2,…,an}、B={b1,b2,…,bm}、
    Figure PCTCN2014095859-appb-100005
    (i=1,2,…,n)、
    Figure PCTCN2014095859-appb-100006
    进行排序的行为关系矩阵MDA0和MDB0
    输出:行为关系矩阵图MDA、MDB中的元素aij(i,j=1,2,…,n)、bij(i,j=1,2,…,m);
    (1)先确定MDA中对角线的元素aii(i=1,2,…,n),依次判断ai(i=1,2,…,n)是否处于环结构中,若ai不在环结构中,那么输出aii=2,执行步骤(2);否则输出aii=0,执行步骤(2);
    (2)再确定ai,i+1、ai+1,i(i=1,2,…,n-1)的值,在网N1中,依次计算ai与ai+1的行为关系,然后将行为关系转化为整数p,输出ai,i+1=ai+1,i=p,执行步骤(3);
    (3)再确定ai,i+2、ai+2,i(i=1,2,…,n-2)的值;若ai,i+1≠ai+1,i+2,输出ai,i+2=ai+2,i=min{ai,i+1,ai+1,i+2};否则,若ai,i+1=ai+1,i+2=1,那么输出ai,i+2=ai+2,i=1;否则,若ai,i+1=ai+1,i+2≠1,那么判断ai与ai+2的行为关系,并转化为行为关系数值q,输出ai,i+2=ai+2,i=q,执行步骤(4);
    (4)同理,确定ai,i+h、ai+h,i(i=1,2,…,n-h)(h=3…,n-1),输出ai,i+h=ai+h,i,直到最后一个元素a1n,算法终止;
    同理根据该行为关系矩阵图中元素的求解算法,计算出MDB中的元素bij(i,j=1,2,…,m),从而得到矩阵MDB
  3. 如权利要求1所述的基于复杂对应系统的用户行为一致性度测量方法,其特征在于,步骤3-2中,所述计算用户模型与预期模型的行为一致性,其一致性度的求解算法为:
    输入:两个工作流网N1=(P1,T1;F1)和N2=(P2,T2;F2),他们的行为关系矩阵MDA0和MDB0,是由步骤3-1步骤中的行为关系矩阵图中元素的求解算法求得;
    输出:一致性度BP
    (1)先根据MDA0与MDB0中变迁集的对应关系,将MDA0和MDB0分别分为p和q个对应集合,将MDA0依次标记为{a1,a2,…am}、{am+1,am+2,…al}…{as+1,…an},执行步骤(2)
    (2)先根据MDA0中第一个与MDB0对应的集合{a1,a2,…am},取MDA0中前m阶方阵记为模块1,并执行步骤(3);
    (3)根据MDA0中第二个与MDB0对应的集合{am+1,am+2,…al},取MDA0中第1→(m)行和(m+1)→(l)列组成的m×(l-m)阶矩阵及其转置矩阵记为模块2,并执行步骤(4);
    (4)仿照之前的步骤,直到MDA0中第p个与MDB0对应的集合{as+1,…an},取MDA0中第1→(m)行和(s+1)→(n)列组成的m×(n-s)阶矩阵及其转置矩阵记为模块p,并执行步骤(5);
    (5)根据MDA0中第二个与MDB0对应的集合{am+1,am+2,…al},取MDA0中第(m+1)→(l)行和(m+1)→(l)列组成的(l-m)阶方阵记为模块p+1,并执行步骤(6);
    (6)仿照步骤(4),将MDA0中(m+1)→(l)行和(s+1)→(n)列组成的(l-m)×(n-s)阶矩阵及其转置矩阵记为模块p+2,并执行步骤(7);
    (7)一直这样进行下去,直到MDA0中第p个与MDB0对应的集合{as+1,…an},取s+1→n行和s+1→n列组成的(n-s)阶方阵记为模块
    Figure PCTCN2014095859-appb-100007
    并执行步骤(8);
    (8)若p=q,同理将MDB0也分解为
    Figure PCTCN2014095859-appb-100008
    对应模块,并标记模块名从1、
    Figure PCTCN2014095859-appb-100009
    则执行步骤(10);否则若p≠q,将MDB0中非重复对应关系也分解为
    Figure PCTCN2014095859-appb-100010
    对应模块,执行步骤(9);
    (9)锁定重复对应的变迁集合,以重复对应的集合组成的区域依次记为模块
    Figure PCTCN2014095859-appb-100011
    执行步骤(10);
    (10)在MDA0中,依次对模块1、2…,
    Figure PCTCN2014095859-appb-100012
    中的矩阵元素进行排查,找出其ai、ai与MDB0相同模块中不同的元素bi、bj,若p=q,输出一致性度BP,算法终止;否则若p≠q,则锁定模块1c、2c…,(q-p)c,输出一致性度BP,算法终止。
PCT/CN2014/095859 2014-07-10 2014-12-31 基于复杂对应系统的用户行为一致性度测量方法 WO2016004744A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
US15/325,184 US20170169350A1 (en) 2014-07-10 2014-12-31 Method for measuring user behavior consistency degree based on complex correspondence system
DE112014006799.7T DE112014006799T5 (de) 2014-07-10 2014-12-31 Verfahren zur Messung eines Benutzerverhalten-Konsistenzgrades auf Basis eines komplexen Korrespondenzsystems
AU2017100012A AU2017100012A4 (en) 2014-07-10 2017-01-05 Method for measuring user behavior consistency degree based on complex correspondence system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410327709.2 2014-07-10
CN201410327709.2A CN104133808B (zh) 2014-07-10 2014-07-10 基于复杂对应系统的用户行为一致性度测量方法

Related Child Applications (1)

Application Number Title Priority Date Filing Date
AU2017100012A Division AU2017100012A4 (en) 2014-07-10 2017-01-05 Method for measuring user behavior consistency degree based on complex correspondence system

Publications (1)

Publication Number Publication Date
WO2016004744A1 true WO2016004744A1 (zh) 2016-01-14

Family

ID=51806487

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/095859 WO2016004744A1 (zh) 2014-07-10 2014-12-31 基于复杂对应系统的用户行为一致性度测量方法

Country Status (4)

Country Link
US (1) US20170169350A1 (zh)
CN (1) CN104133808B (zh)
DE (1) DE112014006799T5 (zh)
WO (1) WO2016004744A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104133808B (zh) * 2014-07-10 2017-06-13 同济大学 基于复杂对应系统的用户行为一致性度测量方法
CN104281674B (zh) * 2014-09-29 2017-07-11 同济大学 一种基于集聚系数的自适应聚类方法及系统
CN105303047B (zh) * 2015-10-30 2018-03-02 同济大学 一种基于分支进程的模型一致性分析方法及系统
CN111242593B (zh) * 2020-01-09 2022-05-31 东华大学 基于伙伴矩阵的交易系统的重叠对应行为一致性检测方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120137367A1 (en) * 2009-11-06 2012-05-31 Cataphora, Inc. Continuous anomaly detection based on behavior modeling and heterogeneous information analysis
CN103559588A (zh) * 2013-11-15 2014-02-05 安徽理工大学 基于Petri网行为轮廓的日志挖掘方法
CN104133808A (zh) * 2014-07-10 2014-11-05 同济大学 基于复杂对应系统的用户行为一致性度测量方法

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100574326C (zh) * 2007-08-08 2009-12-23 南京大学 基于行为一致条件下进行服务发现和替换方法
CN103679484A (zh) * 2014-01-09 2014-03-26 安徽理工大学 基于行为Petri网的电子商务一致性分析新方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120137367A1 (en) * 2009-11-06 2012-05-31 Cataphora, Inc. Continuous anomaly detection based on behavior modeling and heterogeneous information analysis
CN103559588A (zh) * 2013-11-15 2014-02-05 安徽理工大学 基于Petri网行为轮廓的日志挖掘方法
CN104133808A (zh) * 2014-07-10 2014-11-05 同济大学 基于复杂对应系统的用户行为一致性度测量方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WANG, MIMI ET AL.: "Consistency analysis of behavioral profiles based on multi-sets of transitions of Petri net", COMPUTER ENGINEERING AND DESIGN, vol. 34, no. 3, 16 March 2013 (2013-03-16), pages 935 - 938 *

Also Published As

Publication number Publication date
CN104133808A (zh) 2014-11-05
US20170169350A1 (en) 2017-06-15
CN104133808B (zh) 2017-06-13
DE112014006799T5 (de) 2017-05-11

Similar Documents

Publication Publication Date Title
Wang et al. Linkage based face clustering via graph convolution network
Sun et al. Probabilistic updating of building models using incomplete modal data
Bilalli et al. On the predictive power of meta-features in OpenML
Trappey et al. A patent quality analysis for innovative technology and product development
Sánchez-González et al. Quality indicators for business process models from a gateway complexity perspective
Zunino et al. Commodity predictability analysis with a permutation information theory approach
Nsoesie et al. A simulation optimization approach to epidemic forecasting
CN102930495B (zh) 一种基于隐写测评的隐写分析方法
CN107133257A (zh) 一种基于中心连通子图的相似实体识别方法及系统
WO2016004744A1 (zh) 基于复杂对应系统的用户行为一致性度测量方法
Sabariah et al. Early detection of type II Diabetes Mellitus with random forest and classification and regression tree (CART)
Li et al. Explain graph neural networks to understand weighted graph features in node classification
Gupta et al. A novel strategy for link prediction in social networks
CN108446841A (zh) 一种利用灰色关联确定事故因素层次结构的系统方法
Chen et al. Framework based on communicability to measure the similarity of nodes in complex networks
Hudaib et al. Software Reusability classification and predication using self-organizing map (SOM)
Dong Application of Big Data Mining Technology in Blockchain Computing
CN106778252B (zh) 基于粗糙集理论与waode算法的入侵检测方法
Shafik et al. Boosting nonlinear additive autoregressive time series
CN111124489A (zh) 一种基于bp神经网络的软件功能点数估算方法
Käärik et al. Modeling dropouts by conditional distribution, a copula-based approach
WO2017071127A1 (zh) 一种基于分支进程的模型一致性分析方法及系统
AU2017100012A4 (en) Method for measuring user behavior consistency degree based on complex correspondence system
CN110119966A (zh) 银行客户异常行为分析方法
CN111242593B (zh) 基于伙伴矩阵的交易系统的重叠对应行为一致性检测方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14897357

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 15325184

Country of ref document: US

Ref document number: 112014006799

Country of ref document: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 13.07.2017)

122 Ep: pct application non-entry in european phase

Ref document number: 14897357

Country of ref document: EP

Kind code of ref document: A1