Behavior prediction method and device based on domain knowledge fusion

Abstract

The invention provides a behavior prediction method and device based on domain knowledge fusion. The method comprises the following steps: constructing a domain knowledge base mapping knowledge domain; taking the real-time data of the object of the behavior to be predicted as first input information; matching the first input information by the domain knowledge base mapping knowledge domain, and if matching succeeds, outputting the behavior of the object with the behavior to be predicted; otherwise, second information is obtained, the second information is input information of the user or real-time data in the first input information, and when the second information is the input information of the user, the behavior of the object with the behavior to be predicted is determined based on the input information; and when the second information is the real-time data in the first input information, calculating the behavior of the object with the behavior to be predicted, obtaining the latest real-time data of the object with the behavior to be predicted, and taking the calculation result and the latest real-time data of the object with the behavior to be predicted as the first input information.

Description

Technical Field

[0001] The present invention relates to the technical field of military simulation experiments, and in particular to a behavior prediction method and device based on domain knowledge fusion. Background Art

[0002] As the pace of future information warfare continues to accelerate, higher requirements are placed on commanders' battlefield situation analysis and command decision-making capabilities. How to assist commanders in implementing rapid and accurate intention judgment and analysis of targets is important for obtaining information advantages and decision-making advantages and taking the initiative on the battlefield. Through continuous interaction with the command information system, the latest battlefield intelligence information is obtained. Based on the simulation model, real-time and online interactive simulations are performed on the global or local real system, and simulation entities are continuously constructed and corrected, and ultra-real-time simulation deductions are performed. Based on the simulation model, basic data and historical experience, the possible combat intentions and behaviors of enemy targets are judged, and the prediction results of the next moment are generated and fed back to the command information system. The cycle is repeated, so that commanders can "see through" the future and respond in time, and assist commanders in forming combat plans. The parallel simulation system has changed the non-dominant position of the previous simulation system. Its role has changed from passive to active, static to dynamic, and offline to online, making simulation an integral part of the core capabilities of command decision-making.

[0003] For the prediction of test field entity behavior, the common method at this stage is to classify entity behavior by combining domain knowledge base to achieve prediction. The most common existing classification technology is SVM classification. At present, the SVM entity behavior prediction technology based on Grey Wolf Optimization (GWO) algorithm has a good effect on entity behavior prediction. The implementation process of this method is as follows: Figure 1 First, real-time entity data is obtained from the real test site, and matched with the domain knowledge base through interface processing. The matching algorithm adopts the SVM classification algorithm based on GWO, and the historical intelligence data is classified to achieve prediction, and the prediction results are simulated in ultra-real-time.

[0004] Since the gray wolf optimization algorithm randomly initializes the wolf pack, the search process is not very stable, so the chaotic sequence idea is further introduced to make the samples evenly distributed near the samples. The gray wolf algorithm is used to optimally find the penalty factor and kernel function radius of the SVM vector machine, thereby improving the prediction accuracy of the SVM classifier.

[0005] The idea of ​​differential search is introduced into GWO, and the particle swarm optimization algorithm (PSO) is introduced in the iterative process, so that it can save the optimal solution in the evolution process. The position update formula is:

[0006]

[0007] The SVM regression prediction model has two important parameters, namely the penalty coefficient c and the kernel function radius r. The insensitive loss function parameter ε will also directly affect the prediction accuracy value. The larger the ε, the fewer the number of support vectors, the sparser the solution expression, but the accuracy of regression prediction is reduced. The steps of the improved gray wolf algorithm to optimize the support vector machine prediction model parameter optimization process are as follows.

[0008] Step 1: Normalize the sample data. Normalization is to normalize the raw data of different dimensions to a fixed interval. This paper uses a normalization function to normalize the raw data of different dimensions to [0,1].

[0009] Step 2: Divide the training set and prediction set, establish the support vector machine model, and optimize the support vector machine model parameters to find the optimal solution for the parameters [c, g] of the support vector machine regression prediction model.

[0010] Step 3: Evaluate the performance of the support vector machine regression prediction model. If the performance is not ideal, reset the relevant parameters of the particle swarm algorithm and go to step 2 to re-optimize and output the optimal parameters.

[0011] Step 4: Use the optimal parameters [c, g] to train the support vector machine model, use the model for regression prediction of the prediction set samples, and denormalize the prediction results.

[0012] Although the existing technology utilizes the domain knowledge base, it only uses SVM classification. The utilization of the knowledge base is not complete enough. It is unable to incorporate more information into the preconditions of classification. It also ignores the chain relationship between the knowledge in the knowledge base itself and is unable to mine more information for auxiliary decision-making. Summary of the invention

[0013] In order to solve the above technical problems, the present invention proposes a behavior prediction method and device based on domain knowledge fusion, which is used to solve the technical problems of low utilization of knowledge base and less knowledge association in test field entity behavior prediction technology.

[0014] The present invention proposes a behavior prediction method based on domain knowledge fusion, the method comprising:

[0015] Step S1: construct a knowledge graph of a domain knowledge base, wherein the knowledge graph of the domain knowledge base includes multiple knowledge in the form of "entity-relationship-entity" triples, wherein the entities of the knowledge include time elements, events, object subjects of behaviors to be predicted, and action rules; the relationships of the knowledge include association relationships between events and time sequence relationships; and one or more attributes are set for the entities;

[0016] Step S2: real-time data of the object whose behavior is to be predicted is used as first input information;

[0017] Step S3: The first input information is matched by the knowledge graph of the domain knowledge base. If the match is successful, the behavior of the object to be predicted is output, and the method ends; otherwise, the second information is obtained, and the second information is the user's input information or the real-time data in the first input information. When the second information is the user's input information, the behavior of the object to be predicted is determined based on the input information, and the method ends; when the second information is the real-time data in the first input information, the behavior of the object to be predicted is inferred, and the latest real-time data of the object to be predicted is obtained. The inferred result and the latest real-time data of the object to be predicted are used as the first input information, and step S3 is entered.

[0018] Preferably, after determining the behavior of the object to be predicted, the behavior of the object to be predicted is sent to a simulation system, and the simulation system analyzes the intention of the object based on the behavior of the object to be predicted, and displays the behavior and intention of the object.

[0019] Preferably, the domain knowledge base knowledge graph is constructed in the following manner:

[0020] Obtain historical data and test data of the current test task, extract time information, spatial environment information, test task target information, entity objects, and entity behaviors from the historical data and the test data of the current test task, wherein the entity behaviors are the behaviors of the entity objects; use the test task target information as the behavior information; use the entity objects as the entities of the knowledge graph of the domain knowledge base; use the relationship between the entity and the behavior information as the association relationship between the entities; use the time information and the spatial environment information as attributes; form triples and construct an initial knowledge graph;

[0021] The triples in the initial knowledge graph are taken as entities to be fused, and the entity behaviors are taken as behaviors to be fused; the situation where different descriptions appear in the same entity object and / or the same entity behavior is called co-reference, and based on other existing knowledge bases, several triples that can be fused into a triple and several entity behaviors that can be fused into an entity behavior in the initial knowledge graph are determined, and all triples that can be fused into a triple are fused into a triple, and all entity behaviors that can be fused into an entity behavior are fused into an entity behavior, thereby eliminating co-reference; the initial knowledge graph after eliminating co-reference is used as the knowledge graph of the domain knowledge base.

[0022] Preferably, the time type of the time information includes precise time, range time, and fuzzy time. Precise time uses a specific time point to describe the time when an event occurs; range time uses a certain time period to describe the duration from the occurrence of an event to the end or maintaining a certain state, and clarifies the start time and end time; fuzzy time refers to the time when an event occurs is unclear or is within a certain time range; precise time, range time, and fuzzy time are respectively used as attributes of the knowledge graph of the domain knowledge base;

[0023] The combat target information describes the military targets involved in the entity behavior. According to the military decision-making requirements, it is limited to weapons and equipment, and the combat targets are defined as single targets, multiple targets and composite targets. A single target means that the event only involves one type of weapon and equipment, a multiple target means that the event involves multiple equipment, and a composite target means that the military target involved in the event consists of multiple sub-targets. The combat target information is used as an attribute of the knowledge graph of the domain knowledge base.

[0024] The combat mission information describes a brief description of the actions involved in the entity behavior. The combat mission information is also divided into single task and multi-task. A single task is a description of a single action, and a multi-task is a description of each action to form multiple tasks. There is a logical relationship between the tasks. The combat mission information serves as an attribute of the knowledge graph of the domain knowledge base.

[0025] Preferably, the step S3, matching the first input information by the knowledge graph of the domain knowledge base, and outputting the behavior of the object to be predicted if the match is successful, includes:

[0026] The spatial environment information, location information, and attribute information of the object to be predicted are matched in the knowledge graph of the domain knowledge base. After a successful match, the behavior information corresponding to the successfully matched entity in the knowledge graph of the domain knowledge base is used as the behavior of the object to be predicted.

[0027] Preferably, when the second information is real-time data in the first input information, the behavior of the object to be predicted is estimated, including:

[0028] Step S31: taking the real-time data as the data to be classified, the real-time data is a data set generated by the real-time test field; setting the number of clusters K, randomly selecting K data from the data set to be classified as cluster centers, setting the states of the K cluster centers, and the initial values ​​are all changing;

[0029] Step S32: If there is a cluster center state that has changed, proceed to step S33; otherwise, clustering is completed, and the cluster center corresponding to the cluster to which the real-time data belongs is used as the result of the calculation;

[0030] Step S33: Calculate the Euclidean distance from other data in the data set to be classified to the cluster center, and divide it into the nearest center category to obtain the cluster; calculate a new cluster center for each cluster, if the new cluster center is the same as the original cluster center, then set the cluster center state corresponding to the cluster to unchanged, and enter step S32.

[0031] The present invention provides a behavior prediction device based on domain knowledge fusion, the device comprising:

[0032] Knowledge graph construction module: configured to construct a domain knowledge base knowledge graph, wherein the domain knowledge base knowledge graph includes multiple knowledge in the form of "entity-relationship-entity" triples, wherein the entities of the knowledge include time elements, events, object subjects of the behavior to be predicted, and action rules; the relationships of the knowledge include the association relationship between events and the time sequence relationship; and one or more attributes are set for the entity;

[0033] Information acquisition module: configured to use the real-time data of the object whose behavior is to be predicted as the first input information;

[0034] Matching module: configured to match the first input information with the knowledge graph of the domain knowledge base, and if the match is successful, output the behavior of the object of the behavior to be predicted, and the method ends; otherwise, obtain second information, the second information is the user's input information or the real-time data in the first input information, and when the second information is the user's input information, determine the behavior of the object of the behavior to be predicted based on the input information, and the method ends; when the second information is the real-time data in the first input information, infer the behavior of the object of the behavior to be predicted, obtain the latest real-time data of the object of the behavior to be predicted, use the inferred result and the latest real-time data of the object of the behavior to be predicted as the first input information, and trigger the matching module.

[0035] The present invention provides an electronic device, comprising:

[0036] at least one processor; and

[0037] a memory communicatively connected to the at least one processor; wherein,

[0038] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described above.

[0039] The present invention provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the method as described above.

[0040] The present invention realizes the prediction of entity behavior in parallel test field through domain knowledge fusion technology. Starting from the actual needs, this application proposes a test entity behavior prediction method based on domain knowledge fusion to address the problems of low utilization of knowledge base and less knowledge association in the current test field entity behavior prediction technology.

[0041] Based on the data received from various participating entities in the real test field, a knowledge graph is constructed to conduct ultra-real-time simulation and deduction of the enemy and our own combat situation at different time periods in the future, so as to predict key factors such as the enemy's entity's maneuver direction, behavioral intentions, etc.

[0042] The experimental entity behavior prediction method based on domain knowledge fusion proposed in the present invention has at least two advantages over the existing entity behavior prediction technology. First, it uses the domain knowledge graph for data mining, which makes high use of the relevant domain knowledge base and makes the prediction of entity behavior more accurate; second, the knowledge correlation is high. Through knowledge fusion, the relevant knowledge in the graph is interconnected, and the knowledge is in the network, so that the associated knowledge can assist in the division of more refined entity behavior, which is of great significance for predicting the next behavior trend of the entity.

[0043] It should be understood that the contents described in the summary of the invention are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, among which:

[0045] Figure 1 It is a schematic diagram of the SVM entity behavior prediction method based on the gray wolf algorithm optimization in the prior art;

[0046] Figure 2 It is a flowchart of the behavior prediction method based on domain knowledge fusion of the present invention;

[0047] Figure 3 A schematic diagram of the functional composition of the behavior prediction method based on domain knowledge fusion of the present invention;

[0048] Figure 4 It is a schematic diagram of the fusion of domain knowledge of the present invention;

[0049] Figure 5 A schematic diagram of a flow chart of an embodiment of the present invention;

[0050] Figure 6It is a schematic diagram of the structure of the behavior prediction device based on domain knowledge fusion of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0052] Figure 2 FIG. 1 is a schematic diagram of a behavior prediction method based on domain knowledge fusion. It should be understood that in other embodiments, the order of some steps can be interchanged according to actual needs, or some steps can be omitted or deleted. Figure 2 As shown, the method includes:

[0053] Step S1: construct a knowledge graph of a domain knowledge base, wherein the knowledge graph of the domain knowledge base includes multiple knowledge in the form of "entity-relationship-entity" triples, wherein the entities of the knowledge include time elements, events, object subjects of behaviors to be predicted, and action rules; the relationships of the knowledge include association relationships between events and time sequence relationships; and one or more attributes are set for the entities;

[0054] Step S2: real-time data of the object whose behavior is to be predicted is used as first input information;

[0055] Step S3: The first input information is matched by the knowledge graph of the domain knowledge base. If the match is successful, the behavior of the object to be predicted is output, and the method ends; otherwise, the second information is obtained, and the second information is the user's input information or the real-time data in the first input information. When the second information is the user's input information, the behavior of the object to be predicted is determined based on the input information, and the method ends; when the second information is the real-time data in the first input information, the behavior of the object to be predicted is inferred, and the latest real-time data of the object to be predicted is obtained. The inferred result and the latest real-time data of the object to be predicted are used as the first input information, and step S3 is entered.

[0056] Furthermore, after determining the behavior of the object to be predicted, the behavior of the object to be predicted is sent to a simulation system, and the simulation system analyzes the intention of the object based on the behavior of the object to be predicted, and displays the behavior and intention of the object.

[0057] The present invention can be applied to simulation tests on battlefields, using weapons and equipment as test entities, and the test entity behavior prediction method involved. In view of the low utilization of knowledge bases and less knowledge association in the current test field entity behavior prediction technology, the parallel test field entity behavior prediction is realized through the domain knowledge fusion technology. According to the received data of various training entities in the real test field, a knowledge graph is constructed, and the enemy and our combat situation in different periods in the future are simulated and deduced in ultra-real time, so as to predict the key elements such as the enemy entity's maneuver direction and behavioral intention.

[0058] After receiving the real-time data of the real test field entity, the present invention mines the empirical knowledge such as historical tracks and activity rules in the domain knowledge graph, and combines the needs of our force deployment, key defense targets, our action plan, etc. to check whether it can match the empirical data. If the match is successful, the enemy's possible future combat intentions are obtained; if the match is unsuccessful, it is necessary to extrapolate the state of the entity according to the enemy's action rules determined according to experience or historical data in the domain knowledge base, and analyze the enemy's combat intentions according to the latest data, or input the enemy's combat intentions. By carrying out ultra-real-time simulation deduction, the possible results of the enemy's engagement under the current combat intention are obtained, and the prediction results are fed back to the commander through the interface processing module, and after assisting him to optimize the decision, the next round of situation prediction work is carried out iteratively according to the above process. The focus of the test entity behavior prediction method is how to construct a domain knowledge graph, process and fuse the data through the knowledge graph, and construct a more accurate entity behavior prediction model.

[0059] The construction method of the knowledge graph of the domain knowledge base is as follows:

[0060] Obtain historical data and test data of the current test task, extract time information, spatial environment information, test task target information, entity objects, and entity behaviors from the historical data and the test data of the current test task, wherein the entity behaviors are the behaviors of the entity objects; use the test task target information as the behavior information; use the entity objects as the entities of the knowledge graph of the domain knowledge base; use the relationship between the entity and the behavior information as the association relationship between the entities; use the time information and the spatial environment information as attributes; form triples and construct an initial knowledge graph;

[0061] The triples in the initial knowledge graph are taken as entities to be fused, and the entity behaviors are taken as behaviors to be fused; the situation where different descriptions appear in the same entity object and / or the same entity behavior is called co-reference, and based on other existing knowledge bases, several triples that can be fused into a triple and several entity behaviors that can be fused into an entity behavior in the initial knowledge graph are determined, and all triples that can be fused into a triple are fused into a triple, and all entity behaviors that can be fused into an entity behavior are fused into an entity behavior, thereby eliminating co-reference; the initial knowledge graph after eliminating co-reference is used as the knowledge graph of the domain knowledge base.

[0062] Among them, the time type of the time information includes precise time, range-type time, and fuzzy time. Precise time uses a specific time point to describe the time when an event occurs. Range-type time uses a certain time period to describe the duration from the occurrence of an event to the end or maintaining a certain state, and clarifies the start time and end time. Fuzzy time means that the time of an event is unclear or within a certain time range, and generally uses a time range to fuzzily describe it. Precise time, range-type time, and fuzzy time are respectively used as attributes of the knowledge graph of the domain knowledge base.

[0063] The combat target information describes the military targets involved in the entity behavior. According to the military decision-making requirements, it is limited to weapons and equipment, and the combat targets are defined as single targets, multiple targets, and composite targets. A single target means that the event only involves one type of weapon and equipment, a multiple target means that the event involves multiple equipment, and a composite target means that the military target involved in the event consists of multiple sub-targets. The combat target information is used as an attribute of the knowledge graph of the domain knowledge base.

[0064] The combat mission information describes a brief description of the actions involved in the entity behavior, mainly reflecting the relationship between the subject and the object. Similarly, combat mission information is also divided into single tasks and multi-tasks. A single task is a description of a single action, while a multi-task is a description of each action to form multiple tasks. There may be logical relationships such as causal relationships and accompanying relationships between tasks. The combat mission information is used as an attribute of the knowledge graph of the domain knowledge base.

[0065] The topic aggregation information summarized from historical data and other existing knowledge bases reflects the common characteristics of related entity behaviors and summarizes the main purpose of entity behaviors. Compared with event types that describe entity behavior characteristics from a macro level, topic aggregation can focus on entity concerns and reflect the precise information of related entities and related events. It is divided into cross-type entity behavior topics and same-type entity behavior topics. Among them, cross-type entity behavior topics reflect the potential semantic associations between all entity behaviors, and explore the topic associations between multiple entity behavior types from a macro perspective, so as to conduct an overall analysis. The same-type entity behavior topics appropriately focus on the focus, reflecting more detailed differences and deeper semantic associations between the same-type entity behaviors. The topic aggregation information serves as an attribute of the knowledge graph of the domain knowledge base.

[0066] Military entity text descriptions usually involve a large number of professional words, including weapon and equipment names, geographical location names, proper names, military base names, etc. It is difficult for general dictionaries to accurately segment such words. Therefore, it is necessary to build a domain dictionary to assist in text preprocessing. In other words, obtaining historical data and test data of the same type as the current test task, and extracting time information, spatial environment information, and weapon and equipment information from historical data and test data of the same type as the current test task are based on domain dictionaries.

[0067] The domain dictionary is constructed as follows:

[0068] 1) Perform data cleaning on the original entity behavior and other text data to remove irrelevant parts such as punctuation and stop words in the text;

[0069] 2) Use open source Chinese word segmentation tools to perform word segmentation and filter out low-frequency and useless words based on word frequency statistics;

[0070] 3) Use the segmented data as the content of the domain dictionary.

[0071] Furthermore, the original dictionary is expanded, including: incorporating spatial elements and combat target elements of events as new words into the original dictionary; acquiring various weapon and equipment types of entities as new words and incorporating them into the dictionary to complete the construction of the domain dictionary.

[0072] The step S3, matching the first input information with the knowledge graph of the domain knowledge base, and outputting the behavior of the object to be predicted if the match is successful, includes:

[0073] The spatial environment information, location information, and attribute information of the object to be predicted are matched in the knowledge graph of the domain knowledge base. After a successful match, the behavior information corresponding to the successfully matched entity in the knowledge graph of the domain knowledge base is used as the behavior of the object to be predicted.

[0074] When the second information is the real-time data in the first input information, the behavior of the object to be predicted is estimated, including:

[0075] Step S31: taking the real-time data as the data to be classified, the real-time data is a data set generated by the real-time test field; setting the number of clusters K, randomly selecting K data from the data set to be classified as cluster centers, setting the states of the K cluster centers, and the initial values ​​are all changing;

[0076] Step S32: If there is a cluster center state that has changed, proceed to step S33; otherwise, clustering is completed, and the cluster center corresponding to the cluster to which the real-time data belongs is used as the result of the calculation;

[0077] Step S33: Calculate the Euclidean distance from other data in the data set to be classified to the cluster center, and divide it into the nearest center category to obtain the cluster; calculate a new cluster center for each cluster, if the new cluster center is the same as the original cluster center, then set the cluster center state corresponding to the cluster to unchanged, and enter step S32.

[0078] The present invention provides a specific embodiment of a behavior prediction method based on domain knowledge fusion.

[0079] The experimental entity behavior prediction method based on domain knowledge fusion involved in the present invention is used to predict the actions of both parties in the UAV simulation combat.

[0080] After receiving the real-time data of the real test field entity, the parallel test field data processing module mines the empirical knowledge such as historical tracks and activity patterns in the knowledge graph of the UAV field, and combines the needs of the two sides' force deployment, key defense targets, and our action plan to see if it can match the empirical data. If the match is successful, the enemy's possible future combat intentions will be obtained; if the match is unsuccessful, it is necessary to extrapolate the state of the entity according to the enemy's action rules mined in the knowledge graph, and analyze the enemy's combat intentions based on the latest real-time data, or input the enemy's combat intentions. By conducting ultra-real-time simulation and deduction, the possible results of the enemy under the current combat intentions are obtained, and the prediction results are fed back to the commander through the interface processing module.

[0081] The following is a description of the device used to execute the present application. The specific implementation process and technical effects are described above and will not be repeated below.

[0082] Optionally, an embodiment of the present application provides a behavior prediction device based on domain knowledge fusion, the device comprising:

[0083] Knowledge graph construction module: configured to construct a domain knowledge base knowledge graph, wherein the domain knowledge base knowledge graph includes multiple knowledge in the form of "entity-relationship-entity" triples, wherein the entities of the knowledge include time elements, events, object subjects of the behavior to be predicted, and action rules; the relationships of the knowledge include the association relationship between events and the time sequence relationship; and one or more attributes are set for the entity;

[0084] Information acquisition module: configured to use the real-time data of the object whose behavior is to be predicted as the first input information;

[0085] Matching module: configured to match the first input information with the knowledge graph of the domain knowledge base, and if the match is successful, output the behavior of the object of the behavior to be predicted, and the method ends; otherwise, obtain second information, the second information is the user's input information or the real-time data in the first input information, and when the second information is the user's input information, determine the behavior of the object of the behavior to be predicted based on the input information, and the method ends; when the second information is the real-time data in the first input information, infer the behavior of the object of the behavior to be predicted, obtain the latest real-time data of the object of the behavior to be predicted, use the inferred result and the latest real-time data of the object of the behavior to be predicted as the first input information, and trigger the matching module.

[0086] The above-mentioned device is used to execute the method provided by the aforementioned embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0087] The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital singnal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0088] The above modules can be connected or communicated with each other via a wired connection or a wireless connection. The wired connection may include a metal cable, an optical cable, a hybrid cable, etc., or any combination thereof. The wireless connection may include a connection in the form of a LAN, a WAN, Bluetooth, a ZigBee, or NFC, or any combination thereof. Two or more modules may be combined into a single module, and any one module may be divided into two or more units. Those skilled in the art will clearly understand that for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application.

[0089] It should be noted that the above modules can be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital singnal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processors that can call program codes. For another example, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0090] The electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the electronic device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, near field communication (NFC) or other technologies. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the housing of the electronic device, or an external keyboard, touchpad or mouse, etc.

[0091] The present application also provides a program product, such as a computer-readable storage medium, including a program, which is used to execute the above method embodiment when executed by a processor.

[0092] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0093] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0094] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0095] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to perform some steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English: Read-Only Memory, abbreviated: ROM), random access memory (English: Random Access Memory, abbreviated: RAM), disk or optical disk and other media that can store program codes.

Claims

1. A behavior prediction method based on domain knowledge fusion, characterized in that: The method comprises the following steps: Step S1: construct a knowledge graph of a domain knowledge base, wherein the knowledge graph of the domain knowledge base includes multiple knowledge in the form of "entity-relationship-entity" triples, wherein the entities of the knowledge include time elements, events, the object subject of the behavior to be predicted, and action rules; the relationships of the knowledge include the association relationship between events and the time sequence relationship; and one or more attributes are set for the entity; Step S2: real-time data of the object whose behavior is to be predicted is used as first input information; Step S3: The first input information is matched by the knowledge graph of the domain knowledge base. If the match is successful, the behavior of the object to be predicted is output, and the method ends; otherwise, the second information is obtained, and the second information is the user's input information or the real-time data in the first input information. When the second information is the user's input information, the behavior of the object to be predicted is determined based on the input information, and the method ends; when the second information is the real-time data in the first input information, the behavior of the object to be predicted is inferred, and the latest real-time data of the object to be predicted is obtained. The inferred result and the latest real-time data of the object to be predicted are used as the first input information, and step S3 is entered.

2. The method according to claim 1, characterized in that After determining the behavior of the object to be predicted, the behavior of the object to be predicted is sent to a simulation system, and the simulation system analyzes the intention of the object based on the behavior of the object to be predicted, and displays the behavior and intention of the object.

3. The method according to claim 1, characterized in that The construction method of the knowledge graph of the domain knowledge base is as follows: Acquire historical data and test data of the current test task, and extract time information, spatial environment information, test task target information, entity objects, and entity behaviors from the historical data and the test data of the current test task, wherein the entity behaviors are behaviors of the entity objects; The test task target information is used as the behavior information; the entity object is used as the entity of the knowledge graph of the domain knowledge base; the relationship between the entity and the behavior information is used as the association relationship between the entities; the time information and the spatial environment information are used as attributes; Form triplets and construct the initial knowledge graph; The triples in the initial knowledge graph are taken as entities to be fused, and the entity behaviors are taken as behaviors to be fused; the situation where different descriptions appear in the same entity object and / or the same entity behavior is called co-reference, and based on other existing knowledge bases, several triples that can be fused into a triple and several entity behaviors that can be fused into an entity behavior in the initial knowledge graph are determined, and all triples that can be fused into a triple are fused into a triple, and all entity behaviors that can be fused into an entity behavior are fused into an entity behavior, thereby eliminating co-reference; the initial knowledge graph after eliminating co-reference is used as the knowledge graph of the domain knowledge base.

4. The method according to claim 3, characterized in that The time types of the time information include precise time, range time, and fuzzy time. Precise time uses a specific time point to describe the time when an event occurs; range time uses a certain time period to describe the duration from the occurrence of an event to the end or maintaining a certain state, and clarifies the start time and end time; fuzzy time refers to the time when an event occurs is unclear or is within a certain time range; precise time, range time, and fuzzy time are respectively used as attributes of the knowledge graph of the domain knowledge base; The combat target information describes the military targets involved in the entity behavior. According to the military decision-making requirements, it is limited to weapons and equipment, and the combat targets are defined as single targets, multiple targets and composite targets. A single target means that the event only involves one type of weapon and equipment, a multiple target means that the event involves multiple equipment, and a composite target means that the military target involved in the event consists of multiple sub-targets. The combat target information is used as an attribute of the knowledge graph of the domain knowledge base. The combat mission information describes a brief description of the actions involved in the entity behavior. The combat mission information is also divided into single task and multi-task. A single task is a description of a single action, and a multi-task is a description of each action to form multiple tasks. There is a logical relationship between the tasks. The combat mission information serves as an attribute of the knowledge graph of the domain knowledge base.

5. The method according to claim 3, characterized in that The step S3, matching the first input information with the knowledge graph of the domain knowledge base, and outputting the behavior of the object to be predicted if the match is successful, includes: The spatial environment information, location information, and attribute information of the object to be predicted are matched in the knowledge graph of the domain knowledge base. After a successful match, the behavior information corresponding to the successfully matched entity in the knowledge graph of the domain knowledge base is used as the behavior of the object to be predicted.

6. The method according to claim 5, characterized in that When the second information is the real-time data in the first input information, the behavior of the object to be predicted is estimated, including: Step S31: taking the real-time data as the data to be classified, the real-time data is a data set generated by the real-time test field; setting the number of clusters K, randomly selecting K data from the data set to be classified as cluster centers, setting the states of the K cluster centers, and the initial values ​​are all changing; Step S32: If there is a cluster center state that has changed, proceed to step S33; otherwise, clustering is completed, and the cluster center corresponding to the cluster to which the real-time data belongs is used as the result of the calculation; Step S33: Calculate the Euclidean distance from other data in the data set to be classified to the cluster center, and divide it into the nearest center category to obtain the cluster; calculate a new cluster center for each cluster, if the new cluster center is the same as the original cluster center, then set the cluster center state corresponding to the cluster to unchanged, and enter step S32.

7. A behavior prediction device based on domain knowledge fusion, characterized in that: The device comprises: Knowledge graph construction module: configured to construct a domain knowledge base knowledge graph, wherein the domain knowledge base knowledge graph includes multiple knowledge in the form of "entity-relationship-entity" triples, wherein the entities of the knowledge include time elements, events, the subject of the object of the behavior to be predicted, and action rules; the relationships of the knowledge include the association relationship between events and the time sequence relationship; and one or more attributes are set for the entity; Information acquisition module: configured to use the real-time data of the object whose behavior is to be predicted as the first input information; Matching module: configured to match the first input information with the knowledge graph of the domain knowledge base, and if the match is successful, output the behavior of the object of the behavior to be predicted, and the method ends; otherwise, obtain second information, the second information is the user's input information or the real-time data in the first input information, and when the second information is the user's input information, determine the behavior of the object of the behavior to be predicted based on the input information, and the method ends; when the second information is the real-time data in the first input information, infer the behavior of the object of the behavior to be predicted, obtain the latest real-time data of the object of the behavior to be predicted, use the inferred result and the latest real-time data of the object of the behavior to be predicted as the first input information, and trigger the matching module.

8. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 6.

Images