A method for mining mobile object behavior patterns based on interaction data

Abstract

The present invention discloses a method for mining mobile object behavior patterns based on interaction data. By analyzing the interaction data between mobile objects and devices, the movement trajectory of people in real space is captured, and candidate behavior patterns are generated through means such as time thresholds. The feasibility of the candidate behavior patterns is calculated using device location information, interaction mode, etc., and finally the corresponding candidate behavior is selected according to the set credibility threshold.

Description

Technical Field

[0001] The present invention relates to the research and application fields of mobile object data mining, and in particular to a mobile object behavior pattern mining method based on interaction data. Background Art

[0002] With the recent proliferation of mobile internet and IoT devices, the movement trajectories of objects and people have been collected, objectively and realistically reflecting the behavior of mobile objects. Therefore, analysis and mining based on trajectory data has enabled refined operations and more intelligent services, creating significant convenience for people. Existing methods for mining mobile object trajectory patterns primarily rely on tracking and capture systems such as cameras and radar to capture the movement trajectories of mobile objects. However, these existing trajectory acquisition methods are not only costly but also pose the risk of leaking user privacy, making them unsuitable for use in private spaces.

[0003] Due to the popularity of electronic devices, people's activity trajectories can be obtained indirectly through the interaction data of electronic devices. This method of acquisition avoids the risk of privacy leakage.

[0004] Therefore, exploring methods for mining mobile object behavior patterns based on device interaction data has important research value and a wide range of applications. Summary of the Invention

[0005] In order to overcome the problem that existing methods are not applicable when user trajectories cannot be directly obtained, the present invention proposes a mobile object behavior pattern mining method based on device interaction data. By analyzing the interaction data between mobile objects and devices, the movement trajectory of people in real space is captured, and candidate behavior patterns are generated through means such as time thresholds. The feasibility of the candidate behavior patterns is calculated using device location information, interaction methods, etc., and finally the corresponding candidate behaviors are selected according to the set credibility threshold.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for mining mobile object behavior patterns based on interaction data, comprising:

[0008] S1, construct the spatial structure data of the space where the moving object is located into an undirected graph G = (V, E); each subspace is marked as a vertex v (v∈V) of the undirected graph, if two subspaces v i With v j connected, then it forms an edge of the undirected graph (v i , v j )∈E;

[0009] S2, collects the interactive data of the mobile object and stores the interactive behavior of the mobile object (d, t), where d is the interactive behavior associated device of the mobile object and t represents the time point when the interactive behavior occurs;

[0010] S3, collect the interactive devices in the space, record them as a, and record the space v and position reliability c of device d, that is, the position reliability of each device is C d ;

[0011] S4, generating candidate behavior patterns based on the time series of interaction record intervals;

[0012] S5, calculate the behavior sequence based on the undirected graph G and spatial information Spatial credibility

[0013] S6: Calculate the event credibility based on the device information involved in the behavior sequence l

[0014] S7, based on time credibility Spatial credibility Event credibility Calculate the credibility of the behavior sequence C l ;

[0015] S8, according to the set credibility threshold δ, filter the behavior sequence that meets C l >δ, as the final behavior pattern.

[0016] In some embodiments, the undirected graph G=(V, E) in step S1 is constructed as follows: each subspace is identified as a vertex v (v∈V) of the undirected graph. If two subspaces V i With V j connected, then it forms an edge of the undirected graph (v i , v j )∈E vertex V and its attribute vertex v attr Connected, attributes include: space area, space temperature.

[0017] In some embodiments, the interactive behavior of moving the object in step S2 can be performed through one or more of voice, body touch, and remote control. The correlation between each interactive mode and the device position is marked as r. mode , where r 触控 >r 遥控 >r 语音 .

[0018] In some embodiments, step S3 device a location loc d Reliability is manually set based on whether the device is portable, requires installation, or is carried around.

[0019] In some embodiments, step S4 generates candidate behavior patterns based on the interaction record interval time sequence, including:

[0020] Step S41: Select a time interval sequence T0, T1, T2, ...T n , T0 <T1<T2…≤T n ;

[0021] Step S42: record the interaction sequence (d1, t1), (d2, t2), (d3, t3) ... (d m , t m ), construct a behavior sequence of length 1 to m-1 in chronological order, record {L i |i∈[1,m-1]};

[0022] Step S43: Use time interval T i-1 With T i For set L i All the behavior sequences in the behavior sequence are screened to find all the two consecutive behaviors (d j ,t j ), (d j+1 , t j+1 ) Time interval t j+1 -t j ∈(T i-1 , T i ], that is, the behavior sequence must satisfy The sequence of behaviors that meet the conditions is denoted as

[0023] Step S44: Calculate the behavior sequence set Time credibility

[0024] In some embodiments, the behavior sequence of length 1 to m-1 constructed in step 42 must be connected in the original interaction record, that is, for interaction records (d1, t1), (d2, t2), (d3, t3) ... (d m , t m ) constructs a behavior sequence of length 2 as {[(d i , t i ), (d i+1 , t i+1 )]|i∈[1,m-1]}.

[0025] In some embodiments, the computational sequence Spatial credibility The calculation method is as follows:

[0026] Step S51: Calculate two behaviors <(d i ,t i ), (d i+1 , t i+1 )> occurs, then the spatial sequence on the shortest path, that is, the vertex sequence, is {v1, v2, v3…v h}, the spatial area size sequence is Then the spatial confidence of the two behaviors is

[0027] Step S52: Behavior sequence 1 with a length of q = {(d1, t1), (d2, t2), (d3, t3), ... (d q , t q )}of

[0028] In some embodiments, the method for calculating the event credibility based on the device information in S6 is:

[0029] Step S61: Event credibility of a single behavior

[0030] Step S62: Behavior sequence 1 = {(d1, t1), (d2, t2), (d3, t3) ... (dm, t m )} event credibility

[0031] In some embodiments, the S7 calculates the credibility of the behavior sequence where w t , w s , w e is the time weight, space weight, event weight, σ is the sigmoid function, var l T is the number of behavior sequence 1 in all candidate behavior patterns i The variance of .

[0032] In some embodiments, the weight calculation method is:

[0033] 1) Manually select some behavior sequences as true behavior pattern sequences and set their credibility to 1; randomly generate behavior sequences and set their credibility to 0;

[0034] 2) Use the generated samples as training samples and obtain w through training t , w s , w e , c.

[0035] The potential benefits of the mobile object behavior pattern mining method based on interaction data disclosed in this application include, but are not limited to: in situations where the mobile object's behavior trajectory cannot be directly obtained, the mobile object's behavior trajectory can be indirectly inferred through its interaction with the device, thereby providing users with more refined intelligent services. The final results of this invention can be provided to users in related fields, such as smart home scenario recommendations and smart hotel scenarios, while protecting user privacy and ensuring the availability of intelligent services. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Flowchart of the present invention DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0038] On the contrary, this application covers any alternatives, modifications, equivalents, and solutions made within the spirit and scope of this application as defined by the claims. Furthermore, to facilitate a better understanding of this application, certain specific details are described in detail below in the detailed description of this application. Those skilled in the art will be able to fully understand this application without these details.

[0039] The following will be combined Figure 1 A method for mining mobile object behavior patterns based on interaction data according to an embodiment of the present application is described in detail. It should be noted that the following embodiments are only used to explain the present application and do not constitute a limitation of the present application.

[0040] In the embodiments of the present application, Figure 1 As shown, a method for mining mobile object behavior patterns based on interaction data includes:

[0041] S1, the spatial structure data of the space where the moving object is located is constructed as an undirected graph (G = (V, E); each subspace is identified as a vertex v (v∈V) of the undirected graph. If two subspaces V i With V j connected, then it forms an edge of the undirected graph (v i , v j )∈E;

[0042] S2, collects the interactive data of the mobile object and stores the interactive behavior of the mobile object (d, t), where d is the interactive behavior associated device of the mobile object and t represents the time point when the interactive behavior occurs;

[0043] S3, collect the interactive devices in the space, record them as a, and record the space v and position reliability c of device d, that is, the position reliability of each device is C d ;

[0044] S4, generating candidate behavior patterns based on the interaction record interval time sequence; steps S41 to S44 must be executed.

[0045] Step S41: Select a time interval sequence T0, T1, T2, ...T n , T0 <T1<T2…≤T n ;

[0046] Step S42: record the interaction sequence (d1, t1), (d2, t2), (d3, t3) ... (d m , t m ), construct a behavior sequence of length 1 to m-1 in chronological order, record {L i |i∈[1,m-1]};

[0047] Step S43: Use time interval T i-1 With T i For set L i All the behavior sequences in the behavior sequence are screened to find all the two consecutive behaviors (d j ,t j ), (d j+1 , t j+1 ) Time interval t j+1 -t j ∈(T i-1 , T i ], that is, the behavior sequence must satisfy The sequence of behaviors that meet the conditions is denoted as

[0048] Step S44: Calculate the behavior sequence set Time credibility Time credibility

[0049] S5, calculate the behavior sequence based on the undirected graph G and spatial information Spatial credibility

[0050] S6: Calculate the event credibility based on the device information involved in the behavior sequence l

[0051] S7, based on time credibility Spatial credibility Event credibility Calculate the credibility of the behavior sequence Cl ;

[0052] S8, according to the set credibility threshold δ, which is set manually and δ∈[0,1). The screening behavior sequence meets C l >δ, as the final behavior pattern.

[0053] In some embodiments, the undirected graph G=(V, E) in step S1 is constructed as follows: each subspace is identified as a vertex v (v∈V) of the undirected graph. If two subspaces V i With V j connected, then it forms an edge of the undirected graph (v i , v j )∈E; vertex v and its attribute vertex v attr Connected, attributes include: space area, space temperature.

[0054] In some embodiments, the interactive behavior of moving the object in step S2 can be performed through one or more of voice, body touch, and remote control. The correlation between each interactive mode and the device position is marked as r. mode , where r 触控 >r 遥控 >r 语音 .

[0055] In some embodiments, step S3 device a location loc d Reliability is manually set based on whether the device is portable, requires installation, or is carried around.

[0056] In some embodiments, the behavior sequence of length 1 to m-1 constructed in step 42 must be connected in the original interaction record, that is, for interaction records (d1, t1), (d2, t2), (d3, t3) ... (d m , t m ) constructs a behavior sequence of length 2 as {[(d i , t i ), (d i+1 , t i+1 )]|i∈[1,m-1]}.

[0057] In some embodiments, the computational sequence Spatial credibility The calculation method is as follows:

[0058] Step S51: Calculate two behaviors <(d i ,t i ), (d i+1 , t i+1)> occurs, then the spatial sequence on the shortest path, that is, the vertex sequence, is {v1, v2, v3…v h}, the spatial area size sequence is Then the spatial confidence of the two behaviors is

[0059] Step S52: Behavior sequence 1 with a length of q = {(d1, t1), (d2, t2), (d3, t3) ... (d q , t q )}of

[0060] In some embodiments, the method for calculating the event credibility based on the device information in S6 is:

[0061] Step S61: Event credibility of a single behavior

[0062] Step S62: Behavior sequence 1 = {(d1, t1), (d2, t2), (d3, t3) ... (d m , t m )} event credibility

[0063] In some embodiments, the S7 calculates the credibility of the behavior sequence where w t , w s , w e is the time weight, space weight, event weight, σ is the sigmoid function, var l T is the number of behavior sequence 1 in all candidate behavior patterns i The variance of .

[0064] In some embodiments, the weight calculation method is:

[0065] 1) Manually select some behavior sequences as true behavior pattern sequences and set their credibility to 1; randomly generate behavior sequences and set their credibility to 0;

[0066] 2) Use the generated samples as training samples and obtain w through training t , w s , w e , c.

[0067] The overall structure of the present invention is as follows Figure 1 As shown, the principle is:

[0068] First, the user behavior sequence is filtered according to the time interval to generate a candidate set of behavior patterns. In the present invention, the behavior trajectory of the mobile object is an ordered sequence S consisting of a set of interactive behavior records, where the user interaction point point = (d, t), d represents the interactive device, and t represents the specific time point of the interactive behavior of the mobile object. Table 1 shows a set of behavior trajectory sets D, which includes a total of four interactive behavior trajectories. The candidate behavior pattern is an ordered sequence P consisting of a set of interactive behavior records, and P satisfies: the maximum interval t between the interactive behaviors of the mobile object i+1 -t i ∈(T i , T i+1 ], where T i Belong to a pre-set time interval sequence T, and T i <T 1+1 .

[0069] Table 1

[0070]

[0071] Assume T i =50s,T i+1 = 150s, the behavior trajectory shown in Table 1 can generate candidate behavior patterns as shown in Table 2, so the time feasibility of the candidate behavior patterns in Table 2 is

[0072] Table 2

[0073]

[0074]

[0075] According to the location identification of the device, the above behavior pattern is converted into a spatial vertex sequence. For the candidate behavior pattern with identification number 3 in Table 2, the result after conversion is {(a0, 11:24:44), (a2, 11:25:12), (a1, 11:25:55)}, and the spatial size sequence is {25, 16, 36, 16}. As shown in the graph G constructed based on the spatial connectivity relationship,

[0076]

[0077] The event credibility of the behavior sequence is calculated based on the location reliability and control mode of the device. Taking the candidate behavior pattern in ID 3 in Table 2 as an example, its location reliability sequence is {0.92, 0.93, 0.99}, and the control mode correlation sequence is {0.95, 0.45, 0.65}. The event credibility C event =(0.92*0.95)*(0.93*0.45)*(0.99*0.65)=0.2353.

[0078] Assume that the time weight, space weight, and event weight are known as w t =0.9, w s =100, w e =2, var l =1.4, then the credibility of the candidate behavior sequence is C l =1.4*σ(0.9*0.02+100*0.000347+2*0.2353)=0.8885. Time weight, space weight, and event weight can be obtained using machine learning methods.

[0079] The setting of the credibility threshold refers to conventional behavior patterns. Assuming that the threshold δ=0.5, the above candidate behavior pattern will be considered as one of the final mined results, otherwise it will be discarded.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for mining mobile object behavior patterns based on interaction data, characterized in that: include: S1, construct the spatial structure data of the space where the moving object is located into an undirected graph G = (V, E); each subspace is marked as a vertex v (v∈V) of the undirected graph, if two subspaces v i With v j connected, then it forms an edge of the undirected graph (v i , v j )∈E; S2, collects the interactive data of the mobile object and stores the interactive behavior of the mobile object (d, t), where d is the interactive behavior associated device of the mobile object and t represents the time point when the interactive behavior occurs; S3, collect the interactive devices in the space, record them as a, and record the space v and position reliability c of device d, that is, the position reliability of each device is C d ; S4, generating candidate behavior patterns based on the time series of interaction record intervals; S5, calculate the behavior sequence l based on the undirected graph G and spatial information, Spatial credibility S6: Calculate the event credibility based on the device information involved in the behavior sequence l S7, based on time credibility Spatial credibility Event credibility Calculate the credibility of the behavior sequence C l ; S8, according to the set credibility threshold δ, filter the behavior sequence that meets C l >δ, as the final behavior pattern; Step S4 generates candidate behavior patterns based on the interaction record interval time series, including: Step S41: Select a time interval sequence T0, T1, T2, ...T n ,T0<T1<T2…<T n ; Step S42: record the interaction sequence (d1, t1), (d2, t2), (d3, t3) ... (d m , t m ), construct a behavior sequence of length 1 to m-1 in chronological order, record {L i |i∈[1,m-1]}; Step S43: Use time interval T i-1 With T i For set L i All the behavior sequences in the behavior sequence are screened to find all the two consecutive behaviors (d j , t j ), (d j+1 , t j+1 ) Time interval t j+1 -t j ∈(T i-1 ,T i ], that is, the behavior sequence must satisfy The sequence of behaviors that meet the conditions is recorded as Step S44: Calculate the behavior sequence set Time credibility 2. The method for mining mobile object behavior patterns based on interactive data according to claim 1, characterized in that: The undirected graph G = (V, E) in step S1 is constructed as follows: each subspace is identified as a vertex v (v∈V) of the undirected graph. If two subspaces V i With V j connected, then it forms an edge of the undirected graph (v i ,v j )∈E; vertex v and its attribute vertex v attr Connected, attributes include: space area, space temperature.

3. The method for mining mobile object behavior patterns based on interactive data according to claim 1, characterized in that: Step S2: The interactive behavior of the mobile object can be carried out through one or more ways of interaction, such as voice, body touch, and remote control. The correlation degree between each interactive mode and the device position is marked as r. mode , where r 触控 >r 遥控 >r 语音 .

4. The method for mining mobile object behavior patterns based on interactive data according to claim 1, characterized in that: Step S3: The location of device a d Reliability is manually set based on whether the device is portable, requires installation, or is carried around.

5. The method for mining mobile object behavior patterns based on interactive data according to claim 1, characterized in that: Step 42 constructs a behavior sequence of length 1 to m-1, which must be connected in the original interaction record. That is, for the interaction records (d1, t1), (d2, t2), (d3, t3)… (d m ,t m ) constructs a behavior sequence of length 2 as {[(d i , t i ),(d i+1 , t i+1 )]|i∈[1,m-1]}.

6. The method for mining mobile object behavior patterns based on interactive data according to claim 1, characterized in that: The calculation behavior sequence l, Spatial credibility The calculation method is as follows: Step S51: Calculate two behaviors <(d i , t i ), (d i+1 , t i+1 )> occurs, then the spatial sequence on the shortest path, that is, the vertex sequence, is {v1, v2, v3…v h }, the spatial area size sequence is Then the spatial confidence of the two behaviors is Step S52: Behavior sequence 1 with a length of q = {(d1, t1), (d2, t2), (d3, t3), ... (d q , t q )}of 7. The method for mining mobile object behavior patterns based on interactive data according to claim 3, characterized in that: The method for calculating the event credibility according to the device information in S6 is: Step S61: Event credibility of a single behavior Step S62: Behavior sequence 1 = {(d1, t1), (d2, t2), (d3, t3) ... (d m , t m )} event credibility 8. The method for mining mobile object behavior patterns based on interactive data according to claim 1, characterized in that: S7 calculates the credibility of the behavior sequence where w t , w s , w e is the time weight, space weight, event weight, σ is the sigmoid function, var l T is the number of behavior sequence 1 in all candidate behavior patterns i The variance of .

9. The method for mining mobile object behavior patterns based on interactive data according to claim 1, characterized in that The weight calculation method is: 1) Manually select some behavior sequences as true behavior pattern sequences and set their credibility to 1; randomly generate behavior sequences and set their credibility to 0; 2) Use the generated samples as training samples and obtain w through training t , w s , w e , c.

Images