A method for discovering multi-source infectious events for mobile objects
Through the multi-source infection event discovery method, the infectious behavior triple and sliding window mechanism are used to solve the problem that a single infection source model in the existing technology cannot capture multi-source infection events, and achieve more accurate and complete infection of mobile objects Mode mining.
Patent Information
- Application Number
- CN202210092565.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-01-26
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-01-26
AI Technical Summary
The existing mobile object-oriented infectious event model only considers a single source of infection and cannot effectively capture potential infectious events caused by multiple sources of infection, resulting in the omission of potential infectious behavior.
The multi-source infection event discovery method is adopted, and the infectious behavior triple Eobj=(Sobj, obj, ts) indicates that the mobile object obj is infected by the infection source collection Sobj at the moment. It uses the two-dimensional array of candidate infection sources and the sliding window mechanism to determine Multi-source infection events ensure the accuracy of the infection time of each object and the mining of the largest collection of infection sources.
More potential infection events can be discovered, and multiple infection sources of each uninfected object can be considered through a many-to-one transmission mechanism to ensure the time of the earliest occurrence of infection events and the mining of the largest collection of infection sources, improving the accuracy of the infection pattern and integrity.
Smart Images

Figure CN114490818B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of trajectory data mining, in particular to a method for mining multi-source infectious events oriented to mobile objects. Background Art
[0002] The rapid development of satellite positioning technology, wireless communications, and mobile devices has made it easy and cheap to obtain the spatiotemporal trajectory data of moving objects. It has attracted a large number of researchers to analyze and mine these data in order to extract useful patterns that can be applied to many types of services, including urban computing, location prediction, travel recommendations, and traffic management. It can be seen that trajectory data pattern mining for moving objects has practical research and application value.
[0003] The mining of infectious patterns of mobile objects is attracting widespread attention from researchers. Existing research schemes only give a strict and single definition of the infectious event model: when a normal object o comes into close contact with an infectious source s for a continuous time period ΔT (the distance between o and s is less than the distance threshold d), it is determined that the object o is infected by s. This model uses a one-to-one propagation mechanism, but in real life, a normal object may be infected by multiple infectious sources. Therefore, existing schemes may miss potential infectious events. For example, the Chinese patent application with application publication number: CN112632151A, entitled: A method for mining infectious behaviors of mobile objects based on binary filtering, does not fully consider the situation of multiple infectious sources and has the above-mentioned problems. Summary of the invention
[0004] In order to solve the above problems, the present invention provides a mobile object-oriented multi-source infectious event mining method that can discover more potential infectious events.
[0005] In order to achieve the above object, the present invention is implemented by the following technical solutions:
[0006] The present invention is a method for discovering multi-source infectious events for mobile objects, which records multi-source infectious events as infectious behavior triples E obj =(S obj ,obj,ts), where S is the set of confirmed infection sources, E obj Indicates that the mobile object obj is infected by the source set S at time ts obj Infection. If S obj If it is empty, then obj is the initial infection source object, let O be the set of mobile objects, T be the set of trajectory data sampling time points, and R be the set of mined multi-source infection events; the specific steps are as follows:
[0007] Step 1, assume that O S is the initial source of infection, and O SEach source of infection j According to the infectious behavior triple E sj =(NIL,s j ,1) in the form of set R, and at the same time add s j Add the set S, and execute step 2 for the first time point in the trajectory data sampling time point set T;
[0008] Step 2: At the current time point t, clear the temporary infection source set S t ;
[0009] Step 3: For each uninfected object o in the mobile object set O i Analyze and update i The two-dimensional array set C[i][t] of candidate infection sources at time t;
[0010] Step 4: According to the state of C[i][t], determine whether a multi-source infection event occurs; if it occurs, calculate o i The source of infection set S oi , and construct a new multi-source infection event (S oi ,o i , t) add R, and at the same time add o i Join the temporary infection source set S t ; Otherwise, the sets R and S t remain unchanged;
[0011] Step 5: Set the temporary infection source set S t Add S, and the current time point t is processed;
[0012] Step 6: Determine whether all the trajectory time point set T has been processed. If not, process the next time point and go to step 2. If it has been processed, it indicates that the infectious event mining is completed, and R is the final multi-source infectious pattern mining result set.
[0013] A further improvement of the present invention is that the candidate infection source two-dimensional array set C[i][t] in step 3 is expressed as:
[0014] C[i][t]={o j |o j ∈S∧Dist(o i , o j , t)≤d}
[0015] Among them, S represents the set of currently identified infection sources, Dist(o i ,o j ,t) represents the moving object o i and jThe Euclidean distance at time t, d is the distance threshold of the infectious event, and C[i][t] stores the distance between the mobile object o at time t. i All candidate infection sources whose distance is not greater than the distance threshold d.
[0016] A further improvement of the present invention is that: the condition for determining whether a multi-source infection event occurs in step 4 is: for object o i , if there is a sliding window W in T a (ts=W a .end), a multi-source infection event is determined to have occurred if and only if the following two conditions are met at the same time:
[0017] Condition 1: For W a At any time point t k , there is at least one object s j Belongs to S oi , satisfying Dist(o i ,s j ,t k )≤d, that is:
[0018]
[0019] Condition 2: For any end time in W a Sliding window W before .end b , there is at least one time point t k , any object s j Belongs to S oi , do not meet the distance threshold requirement, that is:
[0020]
[0021] Where S oi For the moving object o i The set of infection sources, d is the infection distance threshold, Dist(o i ,s j ,t k ) represents the object o i and infectious agents j In t k Euclidean distance at time, W a .end indicates the sliding window W a The end timestamp of the .
[0022] A further improvement of the present invention is that: in step 4, the determination operation of the multi-source infection event is: let S oi Indicates o i The set of infection sources is initialized to be empty. In the sliding window time series [t–τ+1, t], the following operations are performed on each time point k in chronological order:
[0023] Step 4.1, if k<1, then exit the loop and step 4 ends; otherwise, execute step 4.2;
[0024] Step 4.2: If the candidate infection source set C[i][k] corresponding to time k is not empty, add C[i][k] to S oi ,implement
[0025] Step 4.3; If C[i][k] is empty, execute step 4.4;
[0026] Step 4.3, determine whether the window time series has been processed. If not, process the next time point and go to step 4.2. If it is processed, it indicates that the moving object o i Determine the infection and construct a multi-source infection event (S oi ,o i ,k) Add the contagion event mining result set R and add o i Join the temporary infection source set S t ; Step 4.4, move object o i If there is no infection at time k, the loop judgment is exited and step 4 is completed.
[0027] The beneficial effects of the present invention are as follows: 1. When mining multi-source infectious behaviors, the present invention first maintains a two-dimensional array set C of candidate infectious sources, and then determines the infectious event based on the state of C to ensure the time of the earliest infectious event, that is, to ensure the accuracy of the time when each object is infected; at the same time, ensure that the largest set of infectious sources is obtained, that is, to mine all infectious sources that infect the object. 2. The multi-source infectious event model proposed by the present invention adopts a many-to-one propagation mechanism, and considers multiple infectious sources for each uninfected object. Therefore, the multi-source infectious pattern mining algorithm can discover more potential infectious events. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 It is a flow chart of the present invention.
[0029] Figure 2 Schematic diagram of multi-source infection model. DETAILED DESCRIPTION
[0030] In order to enhance the understanding of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. The embodiments are only used to explain the present invention and do not limit the protection scope of the present invention.
[0031] The present invention is a method for discovering multi-source infectious events for mobile objects. First, the definition of the sliding window is given: let τ be the width threshold of the sliding window, then the sliding window W aIt is composed of a sequence of τ consecutive sampling time points, W a ={t a ,t a+1 ,…,t a+τ-1}.
[0032] Based on the definition of the sliding window above, the definition of multi-source infection behavior is given as follows: given a distance threshold d, for an uninfected object o i For example, if there is a sliding window W a (ts=W a .end) satisfies the following two conditions, then judge o i Infected at time ts:
[0033] Condition 1: For W a At any time point t k , there is at least one object s j Belongs to S oi , satisfying Dist(o i ,s j ,t k )≤d, that is:
[0034]
[0035] Condition 2: For any end time point at W a .end before the sliding window W b , there is at least one time point t k , any object s j Belongs to S oi , do not meet the distance threshold requirement, that is:
[0036]
[0037] Where Dist(o i ,o j ,t k ) represents the object o i and j In t k Euclidean distance at time, W a .end indicates the sliding window W a The end time point, S oi For the moving object o i We need to find a sliding window with a length of τ and the earliest end time, and for each sampling time point in it, there may be multiple sources of infection s j ∈S oi Satisfy the moving object o i and j The distance between them is not greater than the threshold d.
[0038] Based on the above definition of multi-source infectious behavior, suppose that the multi-source infectious event is represented by the infectious behavior triple as E obj =(S obj ,obj,ts), where S is the set of infection sources. obj Indicates that the mobile object obj is infected by the source set S at time ts obj Infection. If S obj If is empty, obj is the initial infection source object. O is the set of mobile objects, T is the set of trajectory data sampling time points, S is the set of confirmed infection sources, and R is the set of mined infection event results.
[0039] like Figure 1 As shown, the specific excavation steps of the present invention are as follows:
[0040] Step 1, assume that O S is the initial source of infection, and O S Each source of infection j According to the infectious behavior triple E sj =(NIL,s j ,1) in the form of set R, and at the same time add s j Add the set S, and execute step 2 for the first time point in the trajectory data sampling time point set T;
[0041] Step 2: At the current time point t, clear the temporary infection source set S t ;
[0042] Step 3: For each uninfected object o in the mobile object set O i Analyze and update i The set C of candidate infectious sources generated at time point t;
[0043] Let n be the number of objects in the mobile object set O, m be the number of time points in the trajectory data time point set T, and the candidate infection source set C be an n*m two-dimensional array set. i The set of candidate infection sources can be expressed as C[i][t], where C[i][t] is:
[0044] C[i][t]={o j |o j ∈S∧Dist(o i , o j , t)≤d}
[0045] Among them, S represents the set of currently infected objects, Dist(o i ,o j ,t) represents the moving object o iand j The Euclidean distance at time t, d represents the distance threshold of the infectious event. C[i][t] stores the distance between the mobile object o at time t. i All potential sources of infection whose distance is no greater than the threshold d.
[0046] Step 4: According to the above conditions 1 and 2, at time point t, for the uninfected object o i The determination of multi-source infection events based on sliding windows and the determination of infection source sets include the following steps: Let S oi Indicates o i The set of infection sources is initialized to be empty. In the sliding window time series [t–τ+1, t], the following operations are performed on each time point k in chronological order:
[0047] Step 4.1, if k<1, then exit the loop and step 4 ends; otherwise, execute step 4.2;
[0048] Step 4.2: If the candidate infection source set C[i][k] corresponding to time k is not empty, add C[i][k] to S oi ,implement
[0049] Step 4.3; If C[i][k] is empty, execute step 4.4;
[0050] Step 4.3, determine whether the window time series has been processed. If not, process the next time point and go to step 4.2. If it is processed, it indicates that the moving object o i Determine the infection and construct a multi-source infection event (S oi ,o i ,k) Add the contagion event mining result set R and add o i Join the temporary infection source set S t ; Step 4.4, move object o i If there is no infection at time k, the loop judgment is exited and step 4 is completed.
[0051] Step 5: Set the temporary infection source set S t Add S, and the current time point t is processed;
[0052] Step 6: Determine whether all the trajectory time point set T has been processed. If not, process the next time point and go to step 2. If it has been processed, it indicates that the infectious event mining is completed, and R is the final multi-source infectious pattern mining result set.
[0053] In this embodiment, it is assumed that the distance threshold d=10 and the time threshold ΔT=6. Figure 2As shown, the distance between object o1 and a single infection source s at 6 consecutive sampling time points (the time points marked with arrows between o1 and S) is not greater than d, so the multi-source infection model can determine that o1 is infected by s, that is, an infection event ({s},o1,8) has occurred; and object o2 and multiple infection sources s and s' meet the threshold judgment requirements at 6 consecutive sampling time points, so the multi-source infection model can also determine that o2 is infected by s and s', that is, an infection event ({s,s'},o2,7) has occurred. The method of the present invention fully considers multiple infection sources and can discover more potential infection events.
[0054] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent process changes made using the contents of the present invention's specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for discovering multi-source infectious events for mobile objects, characterized by: Multi-source infection events are recorded as infection behavior triples E obj = (S obj , obj, ts), where S obj S, S is the set of confirmed infection sources, E obj Indicates that the mobile object obj is infected by the source set S at time ts obj infect; If Sobj is empty, then obj is the initial infection source object, let O be the set of mobile objects, T be the set of trajectory data sampling time points, and R be the set of mined multi-source infection events; the specific steps are as follows: Step 1: Assume OS is the initial set of infectious sources, add each infectious source sj in OS to the set R in the form of infectious behavior triple Esj = (NIL, sj, 1), and add sj to the set S. For the first time point in the trajectory data sampling time point set T, execute step 2; Step 2: At the current time point t, clear the temporary infection source set S t ; Step 3: For each uninfected object o in the mobile object set O i Analyze and update i The two-dimensional array set C[i][t] of candidate infection sources at time t; Step 4: According to the state of C[i][t], determine whether a multi-source infection event occurs based on the sliding window; If it happens, then find o i The source of infection set S oi , and construct a new multi-source infection event (S oi , o i , t) add R, and at the same time i Join the temporary infection source set S t ; Otherwise, the sets R and S t remain unchanged; Step 5: Set the temporary infection source set S t Add S, and the current time point t is processed; Step 6: Determine whether all the trajectory time point set T has been processed. If not, process the next time point and go to step 2. If it has been processed, it indicates that the infectious event mining is completed, and R is the final multi-source infectious pattern mining result set.
2. The method for mining multi-source infectious events for mobile objects according to claim 1, characterized in that: The candidate infection source two-dimensional array set C[i][t] in step 3 is expressed as: ; Among them, S represents the set of currently identified infection sources, Represents a moving object and At time t, the Euclidean distance, d is the infection event distance threshold, and C[i][t] stores the distance between the mobile object and the mobile object at time t. All candidate infection sources whose distance is not greater than the distance threshold d.
3. The method for mining multi-source infectious events for mobile objects according to claim 1, characterized in that: The conditions for determining whether a multi-source infection event has occurred in step 4 are: , if there is a sliding window W in T a , ts = W a .end, a multi-source infection event is determined to have occurred if and only if the following two conditions are met at the same time: Condition 1: For W a At any time point t k , there is at least one object Belongs to S oi , satisfying Dist(o i , s j ,t k ) ≤ d, that is: ; Condition 2: For any end time in W a Sliding window W before .end b , there is at least one time point t k , any object s j Belongs to S oi , do not meet the distance threshold requirement, that is: ; Where S oi For the moving object o i The set of infection sources, d is the infection distance threshold, Dist(o i , s j , t k ) represents the object o i and infectious agents j In t k Euclidean distance at time, W a .end indicates the sliding window W a The end timestamp of the .
4. The method for mining multi-source infectious events for mobile objects according to claim 3, characterized in that: In step 4, the determination operation of multi-source infection events is: let S oi Indicates o i The set of infection sources is initialized to be empty; in the sliding window time series [t–τ+1, t], the following operations are performed on each time point k in chronological order: Step 4.1, if k < 1, then exit the loop and step 4 ends; otherwise, execute step 4.2; Step 4.2: If the candidate infection source set C[i][k] corresponding to time k is not empty, add C[i][k] to S oi , execute step 4.3 If C[i][k] is empty, execute step 4.4; Step 4.3, determine whether the window time series has been processed. If not, process the next time point and go to step 4.
2. If it is processed, it indicates that the moving object o i Determine the infection and construct a multi-source infection event (S oi , o i ,k) Add the contagion event mining result set R and add o i Join the temporary infection source set S t ; Step 4.4, move object o i If there is no infection at time k, the loop judgment is exited and step 4 is completed.
Citation Information
Patent Citations
Moving object infection behavior mining method based on binary filtering
CN112632151A
Infection mode mining method based on trajectory data
CN112084238A
Management of events and moving objects
US20170010111A1
