A method and system for deduction based on a kinetic propagation model
By constructing a dynamic propagation model and adjusting its parameters, the problem that static knowledge graphs cannot analyze the dynamic process of information events was solved, enabling comprehensive deduction and dynamic display of information events and improving the fitting effect of the propagation process.
Patent Information
- Application Number
- CN202310361626.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-04-07
- Publication Date
- 2025-12-30
- Estimated Expiration
- 2043-04-07
AI Technical Summary
Existing knowledge graphs are static and cannot be analyzed from the perspective of the dynamic process of information event evolution and propagation.
A propagation model is constructed based on the dynamic propagation model, the optimal solution of parameters is calculated, the causal relationship between meta-information events is identified, a knowledge graph of events is constructed, and the graph boundary is updated by adjusting the model parameters to dynamically display the development and evolution process of information events.
It enables comprehensive simulation and dynamic display of information events, improves the fitting effect of the information dissemination process, reduces errors and residuals, and is more consistent with the real information dissemination process.
Smart Images
Figure CN116501885B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of big data and artificial intelligence technology, specifically to a deduction method and system based on a dynamic propagation model. Background Technology
[0002] Information dissemination is the activity in which individuals, organizations, and groups exchange information through symbols and media, conveying information, ideas, attitudes, or sentiments to other individuals or groups in order to bring about corresponding changes.
[0003] In existing technologies, event graphs used for information dissemination analysis are constructed based on past similar information events. They calculate the correlation between information events and the results caused by the dissemination of information events, and statistically analyze the frequency of the results caused by information events, thereby inferring the dissemination of information. Existing event knowledge graphs are static knowledge graphs, meaning they only predict the results of information events and cannot analyze information events from the perspective of the dynamic process of information event evolution and dissemination. Summary of the Invention
[0004] (a) Purpose of application
[0005] In view of this, the purpose of this application is to provide a deduction method and system based on a dynamic propagation model to solve the technical problem that existing knowledge graphs are static knowledge graphs and cannot analyze information events from the perspective of the dynamic process of information event evolution and propagation.
[0006] (II) Technical Solution
[0007] This application discloses a derivation method based on a dynamic propagation model, comprising the following steps:
[0008] S1. Construct a propagation model based on dynamics and calculate the optimal solution for the parameters in the propagation model; obtain the number I of the first propagator I based on the optimal solution. a (t); The propagation model includes the state transitions of the mentor and the mentor to the contactor;
[0009] S2. Identify and extract the causal relationships between multiple meta-information events in the actual data, and construct a knowledge graph of events based on the meta-information events and causal relationships, wherein the meta-information events are the nodes of the knowledge graph of events, and the causal relationships are the boundaries between the meta-information events in the knowledge graph of events;
[0010] S3. Based on the propagation model, adjust the parameters in the propagation model; based on the adjusted parameters in the propagation model, obtain the number I of the second propagator I. m (t), based on the number of first propagators I. a (t) and the number of second propagators I m(t) Update the boundaries in the knowledge graph of the subject matter.
[0011] In one possible implementation, the propagation model includes:
[0012]
[0013] Wherein, U(t) is the number of unknowns on day t of the information event, C(t) is the number of contacts on day t of the information event, I(t) is the number of disseminators on day t of the information event, G(t) is the number of recipients of guidance on day t of the information event, and R(t) is the number of immune individuals on day t of the information event; α is the probability of transitioning from unknowns to contacts; β is the probability of transitioning from contacts to disseminators; θ is the probability of transitioning from disseminators to recipients of guidance; λ is the probability of transitioning from disseminators to immune individuals; γ is the probability of transitioning from recipients of guidance to immune individuals; p is the probability of transitioning from contacts to recipients of guidance; q is the probability of transitioning from recipients of guidance to contacts; and ε is the probability of transitioning from immune individuals to unknowns.
[0014] In one possible implementation, the optimal solution for the parameters in the computational propagation model includes:
[0015] S21. Obtain multiple information sample data, each sample data including the amount of data within a time period;
[0016] S22. Calculate the non-zero equilibrium point and the number of regenerables of the propagation model to determine whether the information event is stable;
[0017] S33. When the condition is determined to be stable, the optimal solution of the parameters is calculated based on the multiple information sample data. The parameters include the transfer probability from an unknown person to a contact person, the transfer probability from a contact person to a transmitter, the transfer probability from a transmitter to a recipient of guidance, the transfer probability from a transmitter to an immune person, the transfer probability from a recipient of guidance to an immune person, the transfer probability from a contact person to a recipient of guidance, the transfer probability from a recipient of guidance to a contact person, and the transfer probability from an immune person to an unknown person.
[0018] In one possible implementation, the step of identifying and extracting the causal relationships between multiple meta-information events in the actual data, and constructing a knowledge graph based on meta-information events and causal relationships, further includes vectorizing the knowledge graph into graph words, calculating the similarity and boundary weights of each node, and merging nodes that can be merged based on the similarity.
[0019] In one possible implementation, the number I based on the first propagator I a (t) and the number of second propagators I m(t) Updating the boundaries in the knowledge graph includes: the number of propagators I in the propagation model is I(t), where t is the number of days the information event occurs, and the boundary weight of the knowledge graph is W(t); the number of the first propagator I under the optimal solution parameters is I a (t), modify the propagation model parameters α, β, θ, λ, γ, p, q or ε to obtain the number of second propagators I. m (t), the update formula for the boundary weight W of the knowledge graph is:
[0020]
[0021] As a second aspect of this application, a deduction system based on a dynamic propagation model is also provided, including a propagation model construction module, a knowledge graph construction module, and a boundary update module; wherein, the propagation model construction module is used to construct a propagation model based on dynamics and calculate the optimal solution of the parameters in the propagation model; and to obtain the number I of the first propagator I based on the optimal solution. a (t); The propagation model includes the state transition from the recipient of guidance to the contactor; The knowledge graph construction module is used to identify and extract the causal relationships between multiple meta-information events in the actual data, and construct a knowledge graph based on the meta-information events and causal relationships, wherein the meta-information events are the nodes of the knowledge graph, and the causal relationships are the boundaries between the meta-information events in the knowledge graph; The boundary update module is used to adjust the parameters in the propagation model based on the propagation model; and to obtain the number I of the second propagator I based on the adjusted parameters in the propagation model. m (t), based on the number of first propagators I. a (t) and the number of second propagators I m (t) Update the boundaries in the knowledge graph of the subject matter.
[0022] In one possible implementation, the propagation model includes:
[0023]
[0024] Wherein, U(t) is the number of unknowns on day t of the information event, C(t) is the number of contacts on day t of the information event, I(t) is the number of disseminators on day t of the information event, G(t) is the number of recipients of guidance on day t of the information event, and R(t) is the number of immune individuals on day t of the information event; α is the probability of transitioning from unknowns to contacts; β is the probability of transitioning from contacts to disseminators; θ is the probability of transitioning from disseminators to recipients of guidance; λ is the probability of transitioning from disseminators to immune individuals; γ is the probability of transitioning from recipients of guidance to immune individuals; p is the probability of transitioning from contacts to recipients of guidance; q is the probability of transitioning from recipients of guidance to contacts; and ε is the probability of transitioning from immune individuals to unknowns.
[0025] In one possible implementation, finding the optimal solution for the parameters in the computational propagation model includes performing the following steps:
[0026] S21. Obtain multiple information sample data, each sample data including the amount of data within a time period;
[0027] S22. Calculate the non-zero equilibrium point and the number of regenerables of the propagation model to determine whether the information event is stable;
[0028] S33. When the condition is determined to be stable, the optimal solution of the parameters is calculated based on the multiple information sample data. The parameters include the transfer probability from an unknown person to a contact person, the transfer probability from a contact person to a transmitter, the transfer probability from a transmitter to a recipient of guidance, the transfer probability from a transmitter to an immune person, the transfer probability from a recipient of guidance to an immune person, the transfer probability from a contact person to a recipient of guidance, the transfer probability from a recipient of guidance to a contact person, and the transfer probability from an immune person to an unknown person.
[0029] In one possible implementation, the step of identifying and extracting the causal relationships between multiple meta-information events in the actual data, and constructing a knowledge graph based on meta-information events and causal relationships, further includes vectorizing the knowledge graph into graph words, calculating the similarity and boundary weights of each node, and merging nodes that can be merged based on the similarity.
[0030] In one possible implementation, the number I based on the first propagator I a (t) and the number of second propagators I m (t) Updating the boundaries in the knowledge graph includes: the number of propagators I in the propagation model is I(t), where t is the number of days the information event occurs, and the boundary weight of the knowledge graph is W(t); the number of the first propagator I under the optimal solution parameters is I a(t), modify the propagation model parameters α, β, θ, λ, γ, p, q or ε to obtain the number of second propagators I. m (t), the update formula for the boundary weight W of the knowledge graph is:
[0031]
[0032] (III) Beneficial Effects
[0033] The propagation model constructed through dynamics updates the boundary weights of the knowledge graph of events. Using the knowledge graph as a carrier, the deduction of propagation laws is refined to the content of information events, realizing a comprehensive deduction of information events and a dynamic display of the development and evolution of information events.
[0034] Other advantages, objectives, and features of this application will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from practice of this application. The objectives and other advantages of this application can be realized and obtained through the following description. Attached Figure Description
[0035] The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain and illustrate this application, and should not be construed as limiting the scope of protection of this application.
[0036] Figure 1 This is the system flowchart for this application;
[0037] Figure 2 This is a diagram of the propagation model structure of this application;
[0038] Figure 3 This is a graph showing the relationship between the number of disseminators of this application (I) and time.
[0039] Figure 4 This is a curve fitting graph of the propagation model of this application to the relationship between the number of propagators I and time;
[0040] Figure 5 This is a graph of similar information events merged in this application;
[0041] Figure 6 This is a system structure diagram of this application;
[0042] The module includes: 1. Propagation model construction module; 2. Event knowledge graph construction module; 3. Boundary update module. Detailed Implementation
[0043] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0044] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0045] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0046] like Figure 1-5 As shown, this embodiment provides a deduction method based on a dynamic propagation model, including the following steps:
[0047] S1. Construct a propagation model based on dynamics and calculate the optimal solution for the parameters in the propagation model; obtain the number I of the first propagator I based on the optimal solution. a (t); The propagation model includes the state transition from receiving guidance to becoming a contactor. The transition from receiving guidance to becoming a contactor is based on the secondary contact with the information over time during the propagation process, reflecting the pattern of secondary information propagation. Since the act of publishing information belongs to the behavior of propagator I in the information propagation model, the content required to construct the knowledge graph is only related to the number of propagators I in the information propagation model. The parameters in the propagation model are the transition probabilities between various states, including all or some of the following: unknown, contactor, propagator, receiving guidance, and immune.
[0048] In some embodiments, the propagation model includes:
[0049]
[0050] Wherein, U represents the unknown, C represents the contact, I represents the spreader, G represents the recipient of guidance, and R represents the immune. U(t) is the number of unknowns on day t of the information event, C(t) is the number of contacts on day t of the information event, I(t) is the number of spreaders on day t of the information event, G(t) is the number of recipients of guidance on day t of the information event, and R(t) is the number of immunes on day t of the information event; α is the probability of transitioning from an unknown to a contact; β is the probability of transitioning from a contact to a spreader; θ is the probability of transitioning from a spreader to a recipient of guidance; λ is the probability of transitioning from a spreader to an immune; γ is the probability of transitioning from a recipient of guidance to an immune; p is the probability of transitioning from a contact to a recipient of guidance; q is the probability of transitioning from a recipient of guidance to a contact; and ε is the probability of transitioning from an immune to an unknown.
[0051] The propagation model introduces a novel state variable, the receiving guide G, further refining the information. Secondly, it incorporates the secondary propagation mechanism from the receiving guide G to the information contact C. This construction method ensures that the propagator states in the model converge during the information decay period. Compared to the Sir, Seir, and UCIR models, the propagation model (UCIGR), by considering the secondary propagation mechanism and refining the active individuals in the information propagation process, achieves a better fit to actual information data and more accurately reflects the real information propagation process. Compared to the Sir model, the propagation model (UCIGR) improves the root mean square error (RMSE) by approximately 60%, the mean absolute error (MAE) by approximately 69%, and the residual sum of squares (RSS) by approximately 84%. Compared to the Seir model, the propagation model (UCIGR) improves the RMSE by approximately 40%, the MAE by approximately 42%, and the residual sum of squares (RSS) by approximately 64%. Compared to the UCIR model, the propagation model (UCIGR) shows an improvement of approximately 27% in root mean square error (RMSE), approximately 36% in mean absolute error (MAE), and approximately 47% in residual sum of squares (RSS). Compared to the traditional SIR model, SEIR model, and the UCIR model (an improvement upon the SIR and SEIR models), the UCIGR model demonstrates significant improvements, as shown in Table 1.
[0052] RMSE MAE RSS SIR 379.952 345.041 <![CDATA[4.1866×10 6 ]]> SEIR 249.451 180.141 <![CDATA[1.8046×10 6 ]]> UCIR 204.900 164.737 <![CDATA[1.2175×10 6 ]]> UCIGR 149.573 104.917 <![CDATA[6.4880×10 5 ]]>
[0053] Table 1
[0054] In some embodiments, the optimal solution for the parameters in the computational propagation model includes:
[0055] S21. Obtain multiple information sample data, each sample data including the amount of data within a time period;
[0056] S22. Calculate the non-zero equilibrium point and the number of regenerables of the propagation model to determine whether the information event is stable;
[0057] S33. When the condition is determined to be stable, the optimal solution of the parameters is calculated based on the multiple information sample data. The parameters include the transfer probability from the known person to the contact person, the transfer probability from the contact person to the spreader, the transfer probability from the spreader to the recipient of guidance, the transfer probability from the spreader to the immune person, the transfer probability from the recipient of guidance to the immune person, the transfer probability from the contact person to the recipient of guidance, the transfer probability from the recipient of guidance to the contact person, and the transfer probability from the immune person to the unknown person.
[0058] Information data using "Dalian, epidemic" as keywords was collected from August 25, 2022 to September 22, 2022, a complete dissemination cycle of 29 days. The collected data included the Weibo poster, Weibo content, posting time, number of reposts, number of comments, number of likes, and post links. After data cleaning and deduplication, removing blank or ambiguous Weibo posts, a total of 16,392 data entries were obtained. The number of Weibo posts is divided by the number of days, as shown in Table 2.
[0059]
[0060] Table 2
[0061] like Figure 3-4 As shown, the propagation model is fitted to the actual data. Fitting involves matching the number of propagators I in the propagation model with the actual microblog data to obtain the optimal solution for the parameters of the propagation model.
[0062] The deduction is based on the relationship between the number of propagators I and the parameters. For example, with initial values set as U≈1, C=0, I≈1, G=0, R=0, the optimal parameter solution is obtained as α=4.097, ε=0.105, β=2.240, p=0.071, q=0.157, θ=1.435, λ=0.101, γ=0.251. By modifying the parameters of the propagation model, the specific values of U(t), C(t), I(t), G(t), and R(t) when U, C, I, G, and R are in different states can be obtained.
[0063] For example, by modifying the parameter α, we can obtain the curve of I change. For example, the values of α include 3.0, 3.5, 4.0, 4.5 and 5.0, and there are 5 curves of I change corresponding to the values of α.
[0064] S2. Identify and extract causal relationships between multiple meta-information events in the actual data, and construct a knowledge graph based on the meta-information events and causal relationships. The meta-information events are nodes in the knowledge graph, and the causal relationships are the boundaries between meta-information events in the knowledge graph. The BERT model is used to identify causal relationships between multiple meta-information events in the actual data, and then the BiLSTM-CRF model is used to extract causal logic and construct the knowledge graph.
[0065] In some embodiments, identifying and extracting causal relationships between multiple meta-information events in actual data, and constructing a knowledge graph based on meta-information events and causal relationships, further includes vectorizing the knowledge graph into words, calculating the similarity and boundary weights of each node, and merging nodes that can be merged based on similarity. The knowledge graph can be represented as Graph = {Nodes, Edges, Transforms}, where Nodes = {n1, n2, ..., nk} are nodes, i.e., the set of meta-information events; Edges = {e1, e2, ..., ek} are edges, i.e., causal relationships, each edge pointing from a cause event to a result event; and Transforms = {t1, t2, ..., tk} are the weights of the edges. Figure 5 In this model, weights are represented by the frequency of entity occurrences, indicating the probability that a causal event will cause a specific outcome event. Relation extraction and entity extraction are fundamental to building a knowledge graph. However, repetitive relation descriptions during relation extraction can lead to redundancy in the knowledge graph. Therefore, after relation extraction, text events need to be vectorized, and entities and relations are filtered through similarity calculations. All raw data is processed using the Word2Vec model to obtain word vectors. The vectors of the words forming the event text and their average value are used as the event vector, as shown in the formula:
[0066]
[0067] Where, w i It is event n i The constituent words of ,w ivec For the corresponding word vector, n ivec For event node n i The vector,
[0068] Furthermore, the cosine similarity of vectors between events is calculated using the following formula:
[0069]
[0070] Where, sin(n) i，j ) represents event node n i With n jSimilarity can be used to merge similar events. This similarity can be determined using a similarity threshold, for example, setting a similarity threshold of 0.8. Events with a similarity threshold greater than 0.8 are then manually checked and calibrated again. Figure 5 The diagram shows a merging graph of similar events, where the frequency of occurrence of related entities represents the boundary weights. Boundary weights are simply lengths, but their length reflects the degree of correlation between two entities; a longer boundary indicates a closer relationship. In real-world data, the more times an entity appears, the closer its relationship to the current event; therefore, the frequency can be used to represent the boundary weights.
[0071] S3. Based on the propagation model, adjust the parameters in the propagation model; based on the adjusted parameters in the propagation model, obtain the number I of the second propagator I. m (t), based on the number of first propagators I. a (t) and the number of second propagators I m (t) Update the boundaries in the knowledge graph to dynamically display the development and evolution of information.
[0072] In some embodiments, the number I based on the first propagator I a (t) and the number of second propagators I m (t) Updating the boundaries in the knowledge graph includes: the number of propagators I in the propagation model is I(t), where t is the number of days the information event occurs, the knowledge graph for each day of the propagation model is Graph(t), and the boundary weights of the knowledge graph are W(t); the number of the first propagator I under the optimal solution parameters is I a (t), by modifying any one of the parameters α, β, θ, λ, γ, p, q, ε in the propagation model, we obtain the number I of the second propagators I. m (t), the update formula for the boundary weight W of the knowledge graph is:
[0073] Among them, W * (t) represents the updated weights.
[0074] The Weibo information is divided into days, and the Weibo content for each day is used to construct a knowledge graph using the second part of the technology. Since posting a Weibo is considered an action of communicator I in the information propagation model, the Weibo content required to construct the knowledge graph is only related to the number of communicator states I in the information propagation model. According to the propagation model, the number of communicators I in the UCIGR model is I(t), where t is the number of days the information event has occurred. Therefore, the graph for each day of the propagation model is Graph(t), and the boundary weights of the knowledge graph are W(t); the number of first communicators I under the optimal solution parameters is I.a (t), by modifying any one of the parameters α, β, θ, λ, γ, p, q, ε in the propagation model, we obtain the number I of the second propagators I. m (t), the update formula for the boundary weight W of the knowledge graph is:
[0075] Among them, W * (t) represents the updated weights.
[0076] Specifically, the process involves: summarizing the knowledge graphs of information events for each day to obtain a knowledge graph of events (Graph); optimizing the knowledge graph of events based on entity similarity to obtain the final inferable knowledge graph of events (Graph*); storing the meta-information event content and boundary weights of the inferable knowledge graph of events (Graph*) in the neo4j database for later visualization system calls. neo4j is a graph database with native graph storage and processing capabilities, and it is currently a mainstream graph database. The neo4j database can be modified by calling an interface library implemented through the Python API. The propagation model parameters can be modified through a Python IDE, and the D3 visualization library is used to dynamically display the inferable knowledge graph of events (Graph*). D3 is a JavaScript function library characterized by independent data transformation and rendering, concise code, and no loss of accuracy during scaling.
[0077] The propagation model constructed through dynamics updates the boundary weights of the knowledge graph of events. Using the knowledge graph as a carrier, the deduction of propagation laws is refined to the content of information events, realizing a comprehensive deduction of information events and a dynamic display of the development and evolution of information.
[0078] like Figure 6 As shown, as a second aspect of this application, a deduction system based on a dynamic propagation model is also provided, including a propagation model construction module 1, a knowledge graph construction module 2, and a boundary update module 3; wherein, the propagation model construction module 1 is used to construct a propagation model based on dynamics and calculate the optimal solution of the parameters in the propagation model; and to obtain the number I of the first propagator I based on the optimal solution. a (t); The propagation model includes the state transition from the recipient of guidance to the contactor; The event knowledge graph construction module 2 is used to identify and extract the causal relationships between multiple meta-information events in the actual data, and construct an event knowledge graph based on the meta-information events and causal relationships, wherein the meta-information events are the nodes of the event knowledge graph, and the causal relationships are the boundaries between the meta-information events in the event knowledge graph; The boundary update module 3 is used to adjust the parameters in the propagation model based on the propagation model; and to obtain the number I of the second propagator I based on the adjusted parameters in the propagation model.m (t), based on the number of first propagators I. a (t) and the number of second propagators I m (t) Update the boundaries in the knowledge graph to dynamically display the development and evolution of information.
[0079] In some embodiments, the propagation model includes:
[0080]
[0081] Wherein, U represents the unknown, C represents the contact, I represents the spreader, G represents the recipient of guidance, and R represents the immune. U(t) is the number of unknowns on day t of the information event, C(t) is the number of contacts on day t of the information event, I(t) is the number of spreaders on day t of the information event, G(t) is the number of recipients of guidance on day t of the information event, and R(t) is the number of immunes on day t of the information event; α is the probability of transitioning from an unknown to a contact; β is the probability of transitioning from a contact to a spreader; θ is the probability of transitioning from a spreader to a recipient of guidance; λ is the probability of transitioning from a spreader to an immune; γ is the probability of transitioning from a recipient of guidance to an immune; p is the probability of transitioning from a contact to a recipient of guidance; q is the probability of transitioning from a recipient of guidance to a contact; and ε is the probability of transitioning from an immune to an unknown.
[0082] In some embodiments, calculating the optimal solution for the parameters in the propagation model includes performing the following steps:
[0083] S21. Obtain multiple information sample data, each sample data including the amount of data within a time period;
[0084] S22. Calculate the non-zero equilibrium point and the number of regenerables of the propagation model to determine whether the information event is stable, where the stability of the information event is to determine whether the current information will erupt.
[0085] S33. When the condition is determined to be stable, the optimal solution of the parameters is calculated based on the multiple information sample data. The parameters include the transfer probability from the known person to the contact person, the transfer probability from the contact person to the spreader, the transfer probability from the spreader to the recipient of guidance, the transfer probability from the spreader to the immune person, the transfer probability from the recipient of guidance to the immune person, the transfer probability from the contact person to the recipient of guidance, the transfer probability from the recipient of guidance to the contact person, and the transfer probability from the immune person to the unknown person.
[0086] In some embodiments, the identification and extraction of causal relationships between multiple meta-information events in actual data, and the construction of a knowledge graph based on meta-information events and causal relationships, further include vectorizing the knowledge graph into graph words, calculating the similarity and boundary weight of each node, and merging nodes that can be merged based on the similarity.
[0087] In some embodiments, the number I based on the first propagator I a (t) and the number of second propagators I m (t) Updating the boundaries in the knowledge graph includes: the number of propagators I in the propagation model is I(t), where t is the number of days the information event occurs, and the boundary weight of the knowledge graph is W(t); the number of the first propagator I under the optimal solution parameters is I a (t), modify the propagation model parameters α, β, θ, λ, γ, p, q or ε to obtain the number of second propagators I. m (t), the update formula for the boundary weight W of the knowledge graph is:
[0088]
[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the spirit and scope of the technical solutions of this application, and all such modifications or substitutions should be covered within the scope of the claims of this application.
Claims
1. A method of deduction based on a dynamic propagation model, characterized in that, The method comprises the following steps: S1, constructing a propagation model based on kinetics, and calculating an optimal solution of parameters in the propagation model; obtaining the number of first propagators I based on the optimal solution ; The propagation model comprises state transformation of the acceptor and the acceptor to the contact, and the propagation model comprises: wherein, Ntis the number of unknowns on day t for the information event, Ctis the number of contacts on day t for the information event, S t is the number of spreaders on day t for the information event, Gtis the number of guides on day t for the information event, Itis the number of immunized on day t for the information event; Pctis the probability of moving from unknowns to contacts; Pctis the probability of moving from contacts to spreaders; Pctis the probability of moving from spreaders to guides; Pctis the probability of moving from spreaders to immunized; Pctis the probability of moving from guides to immunized; Pctis the probability of moving from contacts to guides; Pctis the probability of moving from guides to contacts; Pctis the probability of moving from immunized to unknowns; S2, identify and extract the causal relationship between a plurality of meta-information events in actual data, and construct a matter knowledge graph based on the meta-information events and the causal relationship, wherein the meta-information events are nodes of the matter knowledge graph, and the causal relationship is a boundary between the meta-information events in the matter knowledge graph; The calculation of the optimal solution of the parameters in the propagation model comprises: S21, obtain a plurality of information sample data, each sample data comprising a data volume in a time period; S22, calculate the non-zero equilibrium point and the reproductive number of the propagation model to determine whether the information event is stable; S33, when it is determined that the information event is stable, calculate the optimal solution of the parameters based on the plurality of information sample data, the parameters comprising a transition probability from the unknown to the contact, a transition probability from the contact to the propagator, a transition probability from the propagator to the acceptor, a transition probability from the propagator to the immune, a transition probability from the acceptor to the immune, a transition probability from the contact to the acceptor, a transition probability from the acceptor to the contact, and a transition probability from the immune to the unknown; S3, based on the propagation model, adjusting parameters in the propagation model; based on adjusting parameters in the propagation model, obtaining the number of second propagators I , based on the number of first propagators I and the number of second propagators I updating the boundary in the matter knowledge graph.
2. The method of claim 1, wherein, The identification and extraction of the causal relationship between a plurality of meta-information events in actual data, and the construction of a matter knowledge graph based on the meta-information events and the causal relationship further comprise graph word vectorization of the matter knowledge graph, calculation of the similarity of each node and the boundary weight, and merging of nodes that can be merged based on the similarity.
3. The method of claim 2, wherein, The number of the first propagator I is based on The number of the second propagator I Updating the boundary of the affair knowledge graph includes: the number of the propagator I in the propagation model is , wherein t is the number of days of the information event, and the affair knowledge graph boundary weight is The number of the first propagator I under the optimal solution parameter , the propagation model parameter is modified , , , , , , or The number of the second propagator I is obtained The affair knowledge graph boundary weight is The updating formula is: 。 4. A deduction system based on a dynamic propagation model, characterized in that, The method comprises a propagation model construction module, a matter knowledge graph construction module and a boundary updating module; wherein the propagation model construction module is used for constructing a propagation model based on dynamics and calculating an optimal solution of parameters in the propagation model; the number of first propagators I is obtained based on the optimal solution ; the propagation model comprises state transformation of receiving instructors and receiving instructor transfer to contactors, and the propagation model comprises: wherein, is the number of unknowns on day t for the information event, is the number of contacts on day t for the information event, is the number of spreaders on day t for the information event, is the number of receivers on day t for the information event, is the number of immunizers on day t for the information event; is the transition probability from unknowns to contacts; is the transition probability from contacts to spreaders; is the transition probability from spreaders to receivers; is the transition probability from spreaders to immunizers; is the transition probability from receivers to immunizers; is the transition probability from contacts to receivers; is the transition probability from receivers to contacts; is the transition probability from immunizers to unknowns; the fact knowledge graph construction module is used to identify and extract the causal relationship between a plurality of meta-information events in actual data, and construct a fact knowledge graph based on the meta-information events and the causal relationship, wherein the meta-information event is a node of the fact knowledge graph, and the causal relationship is a boundary between the meta-information events in the fact knowledge graph, and the optimal solution of the parameters in the propagation model includes the following steps: S21, obtaining a plurality of information sample data, each sample data including the amount of data within a time period; S22, calculating the non-zero equilibrium point and the reproductive number of the propagation model to determine whether the information event is stable; S33, when it is determined to be stable, calculating the optimal solution of the parameters based on the plurality of information sample data, the parameters including the transition probability from unknowns to contacts, the transition probability from contacts to spreaders, the transition probability from spreaders to receivers, the transition probability from spreaders to immunizers, the transition probability from receivers to immunizers, the transition probability from contacts to receivers, the transition probability from receivers to contacts, and the transition probability from immunizers to unknowns; the boundary updating module is used to adjust the parameters in the propagation model based on the propagation model, and obtain the number of second spreaders I based on the adjusted parameters in the propagation model , based on the number of first spreaders I and the number of second spreaders I to update the boundary in the fact knowledge graph.
5. The system of claim 4, wherein the system is configured to: The identification and extraction of the causal relationship between a plurality of meta-information events in actual data, and the construction of a matter knowledge graph based on the meta-information events and the causal relationship further comprise graph word vectorization of the matter knowledge graph, calculation of the similarity of each node and the boundary weight, and merging of nodes that can be merged based on the similarity.
6. The system of claim 5, wherein the system is configured to: The number of the first propagator I is based on The number of the second propagator I Updating the boundary of the affair knowledge graph includes: the number of the propagator I in the propagation model is , wherein t is the number of days of the information event, and the affair knowledge graph boundary weight is The number of the first propagator I under the optimal solution parameter , the propagation model parameter is modified , , , , , , or The number of the second propagator I is obtained The affair knowledge graph boundary weight is The updating formula is: 。