A knowledge base construction method and system for solving an intercept resource allocation problem

Abstract

The application discloses a knowledge base construction method for solving an intercept resource allocation problem, and can solve the technical problem of the intercept resource allocation problem, especially the dynamic intercept resource allocation problem. The method comprises the following steps: establishing a mathematical model of an allocation problem to be solved, and extracting a problem characteristic vector group; randomly generating b different intercept resource allocation problem samples; constructing a kind of generating rule for solving the problem and an evaluation function of a target; solving the problem samples respectively by using a kind of strategy generating rule and evaluating, selecting a rule suitable for the problem to construct a "characteristic vector group-adaptive rule" binary tuple; dividing the obtained binary tuple into multiple small classes through classification and clustering operations, converting each small class into a piece of knowledge, and constructing a knowledge base; when a new problem occurs, searching for knowledge in the knowledge base by inputting the characteristic vector group of the new problem to solve the problem; evaluating the advantages and disadvantages of an allocation scheme to determine whether to reclassify, and updating and improving the knowledge base.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a knowledge base construction method and system for solving the problem of interception resource allocation. Background Technology

[0002] Interception resource allocation is a core issue in command and decision-making. It requires determining a rational allocation scheme based on the type and threat level of incoming targets, the type and performance of interception resources, and the battlefield situation to achieve optimal operational results. The interception resource allocation problem has been proven to be NP-complete. Especially as the types and quantities of targets and interception resources increase, the solution time grows exponentially with the problem size, making it difficult to solve using traditional methods such as target programming or evolutionary algorithms with slow convergence speeds, such as genetic algorithms and ant colony algorithms.

[0003] Furthermore, in the target allocation problem for interception resources, even slight changes to the types and quantities of targets, the types and quantities of interception resources, the attack range of interception resources, problem constraints, and problem parameters render the original allocation scheme inapplicable, necessitating a significant investment of time in re-analyzing and resolving the problem. This is particularly problematic in dynamic interception resource allocation problems, where the problem constantly changes, making it unsuitable for time-critical dynamic problems requiring a restart. Summary of the Invention

[0004] In view of this, the present invention provides a knowledge base construction method and system for solving the interception resource allocation problem, which can utilize existing relevant knowledge to solve the interception resource allocation problem, especially the dynamic interception resource allocation problem.

[0005] This invention mainly solves the following technical problems:

[0006] A method for constructing a knowledge base to solve the problem of interception resource allocation, the method comprising the following steps:

[0007] Step S1: Establish the model of the interception resource allocation problem to be solved, extract the features of the interception resource allocation problem, combine all the extracted features to form a feature vector group, and change the feature values ​​in the feature vector group to randomly generate b different interception resource allocation problem samples.

[0008] Step S2: Construct a kind of generation rule for solving the interception resource allocation problem. All a kind of generation rule can generate an allocation scheme for the interception resource allocation problem.

[0009] Step S3: Construct an evaluation function for the target of the interception resource allocation problem. The constructed evaluation function is used to evaluate the allocation scheme of the interception resource allocation problem.

[0010] Step S4: If a knowledge base has not yet been built, use the a-type generation rules built in step S2 to solve each problem sample generated in step S1, obtaining a-type allocation schemes respectively. Evaluate the a-type allocation schemes obtained according to the evaluation function built in step S3. Record the generation rules corresponding to the allocation schemes whose evaluation function values ​​are higher than a set threshold, and define these generation rules as the adaptation rules for this problem sample. Based on the adaptation rules obtained for each problem sample, construct a "feature vector group - adaptation rule" tuple. If a knowledge base has already been built, use the a-type generation rules built in step S2 to solve the newly arrived interception resource allocation problem, and evaluate the a-type allocation schemes obtained according to the evaluation function built in step S3. Record the generation rules corresponding to the allocation schemes whose evaluation function values ​​are higher than a set threshold, and define these generation rules as the adaptation rules for this problem sample. Based on the adaptation rules obtained for each problem sample, construct a "feature vector group - adaptation rule" tuple.

[0011] Step S5: Perform classification and clustering operations on the "feature vector group - adaptation rule" binary pairs constructed in step S4. The classification and clustering operations are as follows: First, classify according to the adaptation rule, classify binary pairs with the same adaptation rule into major categories, and within each major category, perform clustering operation according to the feature vector group, and cluster binary pairs with similar feature vector groups into minor categories. The feature vector range group of the minor category is taken as the range of each feature vector in the minor category, and the cluster center point of the minor category is calculated.

[0012] Step S6: Each subclass in step S5 is formed into a knowledge item, and all the knowledge is used to build a knowledge base. The knowledge is in the form of "feature vector range group - adaptation rule - center point".

[0013] The method for solving the interception resource allocation problem using the above knowledge base involves the following steps:

[0014] Step S7: If a new interception resource allocation problem arises, extract the feature vector set of the new problem and input the feature vector set of the new problem into the knowledge base for retrieval; use the retrieved knowledge to solve the problem and obtain the allocation scheme for the new problem.

[0015] Step S8: Evaluate the allocation scheme obtained by solving the problem using the retrieved knowledge. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful. Update the knowledge according to the tuple of the new problem and update and improve the knowledge base. Proceed to step S7. If the evaluation value is lower than the set evaluation threshold, the knowledge application fails. Proceed to step S4.

[0016] Preferably, in step S1, features of the interception resource allocation problem are extracted. Feature extraction requires establishing a model of the interception resource allocation problem and extracting features of the problem, including extracting features based on the types of targets, the threat level of each type of target, the number of targets, the types of interception resources, the number of each type of interception resource, and the strike range of the interception resources. The extracted features of the problem can characterize the interception resource allocation problem.

[0017] Preferably, in step S1, the method for obtaining the problem sample is as follows: based on the feature vector group of the intercepted resource allocation problem, set the value range of each vector in the feature vector group, and randomly generate a feature vector group within the set value range. Each newly generated feature vector group represents a newly generated problem sample.

[0018] Preferably, in step S3, when evaluating the allocation scheme, a threshold for the evaluation function is set. When the evaluation function value is greater than the set threshold, the allocation scheme is a high-quality scheme, and the generation rule corresponding to the allocation scheme is applicable to solving the interception resource allocation problem. When the evaluation function value is not greater than the set threshold, the allocation scheme is a low-quality scheme, and the generation rule corresponding to the allocation scheme is not applicable to solving the interception resource allocation problem.

[0019] Preferably, in step S4, there may be more than one adaptation rule for the interception resource allocation problem. If the evaluation values ​​of the allocation schemes obtained by multiple generation rules are all higher than the threshold, then all multiple generation rules are recorded, and each generation rule and the problem are respectively constructed into a "feature vector group - adaptation rule"; if the evaluation values ​​of the allocation schemes obtained by solving the problem are higher than r generation rules, then only the r generation rules with higher evaluation values ​​are recorded and a binary tuple is constructed.

[0020] Preferably, in step S5, the binary groups with similar feature vector sets in the major category are clustered into a minor category, the feature vector range set of the minor category is taken as the range of each feature vector in the category, and the center point of the minor category is calculated, including:

[0021] Step S51: Use the K-means algorithm to cluster the binary pairs in the major category based on the feature vector group: find k clusters by iteratively, so that the overall error obtained when using the mean of these k clusters to represent the corresponding sub-category problem samples is minimized, where the number of categories k is selected according to the capacity of the knowledge base;

[0022] Step S52: After the clustering operation, multiple subclasses are obtained. The vector values ​​of each feature vector group in each subclass are extracted, and the distribution range of each vector in the subclass is calculated. Thus, the value range of each feature in the feature vector group of this subclass is obtained, which serves as the criterion for judging whether a new interception resource allocation problem belongs to this subclass.

[0023] Step S53: For each pair i in the subclass, calculate the sum of the Euclidean distances between the feature vector group of pair i and the feature vector groups of all other pairs (pairs other than i), and take the feature vector group of the pair with the smallest sum of Euclidean distances as the center point of this subclass.

[0024] Preferably, step S6, constructing the knowledge base, includes: taking each subclass obtained from the above classification and clustering operations as a knowledge entry, the knowledge entry being in the form of "feature vector range group - adaptation rule - centroid"; wherein, the "feature vector range group" in each knowledge entry is the problem feature range of that subclass, that is, the range group of each feature in that class; the "adaptation rule" in each knowledge entry is the adaptation rule of that subclass; and the "centroid" in each knowledge entry is the centroid of that subclass, that is, the feature vector group of the binary tuple that has the shortest Euclidean distance sum with the feature vector groups of other binary tuples in the subclass; the knowledge represented by all subclasses together constitutes the knowledge base for the interception resource allocation problem;

[0025] Preferably, the retrieval method in step S7 includes:

[0026] Step S71: If, based on the feature vector group of the new interception resource allocation problem, knowledge matching the input is found in the knowledge base, that is, each element in the feature vector group of the new problem is within the corresponding feature range of a certain knowledge, then the adaptation rule in this knowledge is used to solve the new problem;

[0027] The matching process between the feature vector group of the new problem and the knowledge in the knowledge base involves comparing each vector value in the feature vector group of the new problem with the corresponding vector range in the feature range of each knowledge. If all the values ​​in the feature vector group of the new problem belong to the feature range group of a certain knowledge, the matching is successful; otherwise, the matching fails, and the process proceeds to step S72.

[0028] In the knowledge matching process, if multiple pieces of knowledge match the new problem, the Euclidean distance between the feature vector group of the new problem and the feature vector group of the center point of the matched multiple pieces of knowledge is calculated, and the knowledge with the shortest Euclidean distance is taken as the final knowledge that matches it.

[0029] Step S72: If no matching knowledge is found in the knowledge base based on the feature vector set of the new problem, then the knowledge with the closest Euclidean distance to the feature vector set of the new problem is searched for and solved.

[0030] If no match is found in the knowledge base during the knowledge matching process, the knowledge with the closest Euclidean distance to the feature vector group of the new question is searched. This involves calculating the Euclidean distance between the feature vector group of the new question and the feature vector group of the center point of each knowledge in the knowledge base, and taking the knowledge to which the center point with the shortest Euclidean distance belongs as the approximate matching knowledge.

[0031] Preferably, in step S8, the allocation scheme obtained by solving the retrieved knowledge is evaluated and the knowledge base is updated and improved, including:

[0032] Step S81: Evaluate the allocation scheme obtained by using the knowledge according to the set evaluation function, and obtain the evaluation function value of the scheme. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful.

[0033] Step S82: Update the knowledge used above using the corresponding tuples of the new problem:

[0034] The tuples of the new interception resource allocation problem are added to the subclass corresponding to the knowledge, and the knowledge corresponding to the subclass is updated. The knowledge includes feature vector range groups, adaptation rules, and centroids. Since the tuples of the new problem match the subclass, the feature vector range groups and adaptation rules of the subclass remain unchanged after the tuples of the new problem are added to the subclass. However, the centroids need to be recalculated. The calculation process is to calculate the sum of the Euclidean distances of the feature vector groups of each tuple to the feature vector groups of all other tuples, and take the feature vector group of the tuple with the smallest sum of Euclidean distances as the centroid of the subclass.

[0035] A knowledge base construction system for solving the problem of interception resource allocation includes a data acquisition module, an initialization module, a problem-solving module, a classification and clustering module, a construction module, a monitoring module, an application module, and a maintenance module.

[0036] Data acquisition module: Configured to build a model of the interception resource allocation problem to be solved, extract the features of the interception resource allocation problem, and combine all the extracted features to form a feature vector group; change the feature values ​​in the formed feature vector group to randomly generate b different interception resource allocation problem samples;

[0037] Initialization module: configured to construct a generation rule for solving the interception resource allocation problem, and to construct an evaluation function for the objective of the interception resource allocation problem;

[0038] Problem Solving Module: Configured to solve the problem samples in the data acquisition module by solving the a-type scheme generation rules in the initialization module to obtain a-type allocation schemes, and evaluate them according to the evaluation function. Record the scheme generation rules corresponding to the allocation schemes with evaluation values ​​higher than the threshold, and define them as the adaptation rules of the problem sample. Construct a "feature vector group - adaptation rule" tuple based on the adaptation rules of the problem sample.

[0039] The classification and clustering module is configured to perform classification and clustering operations on the "feature vector set - adaptation rule" pairs constructed by the problem-solving module. The classification and clustering operations are as follows: First, classify according to the adaptation rule, classify pairs with the same adaptation rule into major categories. Within each major category, perform clustering operation according to the feature vector set, and cluster pairs with similar feature vector sets into minor categories. The feature vector range set of the minor category is taken as the range of each feature vector in the minor category, and the cluster center point of the minor category is calculated.

[0040] The building module is configured to form each subclass as a knowledge item and use all the knowledge to build a knowledge base. The knowledge is in the form of "feature vector range group - adaptation rule - center point".

[0041] Listening module: Configured to listen for new intercepted resource allocation issues; if a new issue arises, trigger the application module.

[0042] Application module: configured to extract the feature vector set of the new interception resource allocation problem, input it into the knowledge base for retrieval, and use the retrieved knowledge to solve the problem and obtain the allocation scheme of the problem;

[0043] Maintenance module: Configured to evaluate the allocation scheme obtained by solving the knowledge retrieved by the application module. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful. The knowledge to which it belongs is updated according to the tuple of the new problem, and the knowledge base is updated and improved, triggering the monitoring module. If the evaluation value is lower than the set evaluation threshold, the knowledge application fails, triggering the problem solving module.

[0044] The present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it implements the knowledge base construction method for solving the interception resource allocation problem as described above.

[0045] Beneficial effects

[0046] (1) This invention addresses the problem of interception resource allocation, especially dynamic large-scale interception resource allocation, which is difficult to solve directly due to the time constraints of the problem. It realizes the use of knowledge to solve the problem, thereby quickly obtaining a better quality allocation scheme.

[0047] (2) Considering the limited capacity of the knowledge base, this invention extracts the feature vector group of the allocation problem to represent the problem and connects the feature vector group with its generation rule instead of directly storing its allocation scheme. This makes it easier to retrieve the feature vector group in the knowledge base and greatly reduces the storage space occupied by the knowledge.

[0048] (3) The present invention uses classification and clustering operations to highly condense the previous solution experience into knowledge and store it in the knowledge base, which greatly reduces the amount of data that needs to be stored. This overcomes the shortcomings of knowledge being difficult to store and directly use, knowledge incompleteness due to capacity limitations of the knowledge base, and difficulty in searching for problems, and improves the generalization ability of the knowledge base.

[0049] (4) This invention makes full use of the previous results of solving resource allocation problems, summarizes the experience of solving different problems to build a knowledge base, and identifies the feature vector group of new problems to match the generation rules, which saves a lot of time and computation costs. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0051] Figure 2 This is a schematic diagram illustrating the problem feature extraction of the present invention;

[0052] Figure 3 This is a schematic diagram of the feature vector group of the problem in this invention;

[0053] Figure 4 This is a diagram of the clustering operation based on the classification operation of the present invention;

[0054] Figure 5 This is a schematic diagram of the knowledge base structure of the present invention. Detailed Implementation

[0055] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0056] like Figure 1 As shown, the present invention provides a knowledge base construction method for solving the interception resource allocation problem, comprising the following steps:

[0057] Step S1: As Figure 2 As shown, a model for the interception resource allocation problem is established, features of the interception resource allocation problem are extracted, and all extracted features are combined to form a feature vector group. The feature values ​​in the formed feature vector group are changed to randomly generate b different interception resource allocation problem samples, such as... Figure 3 As shown;

[0058] Step S2: Construct a kind of generation rule for solving the interception resource allocation problem. All a kind of generation rule can generate an allocation scheme for the interception resource allocation problem.

[0059] Step S3: Construct an evaluation function for the target of the interception resource allocation problem. The constructed evaluation function is used to evaluate the allocation scheme of the interception resource allocation problem.

[0060] Step S4: If a knowledge base has not yet been built, use the a-type generation rules built in step S2 to solve each problem sample generated in step S1, obtaining a-type allocation schemes respectively. Evaluate the a-type allocation schemes obtained according to the evaluation function built in step S3. Record the generation rules corresponding to the allocation schemes whose evaluation function values ​​are higher than a set threshold, and define these generation rules as the adaptation rules for this problem sample. Based on the adaptation rules obtained for each problem sample, construct a "feature vector group - adaptation rule" tuple. If a knowledge base has already been built, use the a-type generation rules built in step S2 to solve the newly arrived interception resource allocation problem, and evaluate the a-type allocation schemes obtained according to the evaluation function built in step S3. Record the generation rules corresponding to the allocation schemes whose evaluation function values ​​are higher than a set threshold, and define these generation rules as the adaptation rules for this problem sample. Based on the adaptation rules obtained for each problem sample, construct a "feature vector group - adaptation rule" tuple.

[0061] Step S5: Perform classification and clustering operations on the "feature vector group - adaptation rule" binary pairs constructed in Step S4. Specifically, the classification and clustering operations are as follows: first, classify according to the adaptation rules, grouping binary pairs with the same adaptation rule into larger categories, such as... Figure 4 As shown, within each major category, clustering is performed based on the feature vector set. Binaries with similar feature vector sets are clustered into smaller categories. The feature vector range set of a smaller category is taken as the range of each feature vector in that smaller category, and the cluster center point of the smaller category is calculated.

[0062] Step S6: As Figure 5 As shown, each subclass in step S5 is formed into a knowledge item, and all the knowledge is used to build a knowledge base. The knowledge is in the form of "feature vector range group - adaptation rule - center point".

[0063] The method for solving the interception resource allocation problem using the above knowledge base involves the following steps:

[0064] Step S7: If a new interception resource allocation problem arises, extract the feature vector set of the new problem and input the feature vector set of the new problem into the knowledge base for retrieval; use the retrieved knowledge to solve the problem and obtain the allocation scheme for the new problem;

[0065] Step S8: Evaluate the allocation scheme obtained by solving the problem using the retrieved knowledge. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful. Update the knowledge according to the tuple of the new problem and update and improve the knowledge base. Proceed to step S7. If the evaluation value is lower than the set evaluation threshold, the knowledge application fails. Proceed to step S4.

[0066] Step S1 involves extracting features from the interception resource allocation problem. Feature extraction requires establishing a model of the interception resource allocation problem. Extracted features include the types of targets, the threat level of each target, the number of targets, the types of interception resources, the quantity of each type of interception resource, and the strike range of the interception resources. Examples of features include: the ratio of interception resources to targets is 1 / 3; targets with high threat levels account for 45% of the total number of targets; and targets with a probability of being successfully destroyed by interception resources exceeding 0.9 account for 70%. The extracted features characterize the interception resource allocation problem.

[0067] The method for obtaining the problem sample in step S1 is as follows: based on the feature vector group of the intercepted resource allocation problem, set the value range of each vector in the feature vector group, and randomly generate a feature vector group within the set value range. Each newly generated feature vector group represents a newly generated problem sample.

[0068] In step S2, the rule for generating the solution to the interception resource allocation problem is as follows:

[0069] Rule 1 for scheme generation: The more target threats a certain interception resource can eliminate by being used to combat a certain target, the higher its priority for intercepting that target;

[0070] Rule 2 for scheme generation: If any available interception resources are allocated to a target, then the threat value of the corresponding target is reduced;

[0071] Rule 3 for scheme generation: If the marginal loss caused by deleting a redundant "intercept resource - target" allocation scheme chain is small, the higher its priority will be considered in the allocation process;

[0072] In step S3, when evaluating the allocation scheme, a threshold for the evaluation function is set. When the evaluation function value is greater than the set threshold, the allocation scheme is a high-quality scheme, and the generation rule corresponding to the allocation scheme is applicable to solving the interception resource allocation problem. When the evaluation function value is not greater than the set threshold, the allocation scheme is a low-quality scheme, and the generation rule corresponding to the allocation scheme is not applicable to solving the interception resource allocation problem.

[0073] In step S4, there may be more than one adaptation rule for the interception resource allocation problem. If the evaluation values ​​of the allocation schemes obtained by multiple generation rules are all higher than the threshold, then all multiple generation rules are recorded, and each generation rule and the problem are respectively constructed into a "feature vector group - adaptation rule"; if the evaluation values ​​of the allocation schemes obtained by solving the problem are higher than r generation rules, then only the r generation rules with higher evaluation values ​​are recorded and a binary tuple is constructed.

[0074] Step S5 involves clustering binary pairs with similar feature vector groups within a larger class into a smaller class. The feature vector range of this smaller class is taken as the range of each feature vector in that class, and the center point of this smaller class is calculated, including:

[0075] Step S51: Use the K-means algorithm to cluster the binary pairs in the major category based on the feature vector group: find k clusters by iteratively, so that the overall error obtained when using the mean of these k clusters to represent the corresponding sub-category problem samples is minimized, where the number of categories k is selected according to the capacity of the knowledge base;

[0076] Step S52: After the clustering operation, multiple subclasses are obtained. The vector values ​​of each feature vector group in each subclass are extracted, and the distribution range of each vector in the subclass is calculated. Thus, the value range of each feature in the feature vector group of this subclass is obtained, which serves as the criterion for judging whether a new interception resource allocation problem belongs to this subclass.

[0077] For example, if the values ​​of feature 1 for all pairs in a certain subclass are 10, 12, 11.5 and 14, then the range of values ​​for this feature in the subclass is [10, 14].

[0078] Step S53: For each pair i in the subclass, calculate the sum of the Euclidean distances between the feature vector group of pair i and the feature vector groups of all other pairs (pairs other than i), and take the feature vector group of the pair with the smallest sum of Euclidean distances as the center point of this subclass.

[0079] Step S6, constructing the knowledge base, includes: taking each subclass obtained from the above classification and clustering operations as a knowledge entry, with the knowledge in the form of "feature vector range group - adaptation rule - centroid"; wherein, the "feature vector range group" in each knowledge entry is the problem feature range of that subclass, that is, the range group of each feature in that class; the "adaptation rule" in each knowledge entry is the adaptation rule of that subclass; and the "centroid" in each knowledge entry is the centroid of that subclass, that is, the feature vector group of the binary tuple that has the shortest Euclidean distance sum with the feature vector groups of other binary tuples in the subclass; the knowledge represented by all subclasses together constitutes the knowledge base for the interception resource allocation problem;

[0080] The retrieval method in step S7 includes:

[0081] Step S71: If, based on the feature vector group of the new interception resource allocation problem, knowledge matching the input is found in the knowledge base, that is, each element in the feature vector group of the new problem is within the corresponding feature range of a certain knowledge, then the adaptation rule in this knowledge is used to solve the new problem;

[0082] The matching process between the feature vector group of the new problem and the knowledge in the knowledge base involves comparing each vector value in the feature vector group of the new problem with the corresponding vector range in the feature range of each knowledge. If all the values ​​in the feature vector group of the new problem belong to the feature range group of a certain knowledge, the matching is successful; otherwise, the matching fails, and the process proceeds to step S72.

[0083] In the knowledge matching process, if multiple pieces of knowledge match the new problem, the Euclidean distance between the feature vector group of the new problem and the feature vector group of the center point of the matched multiple pieces of knowledge is calculated, and the knowledge with the shortest Euclidean distance is taken as the final knowledge that matches it.

[0084] Step S72: If no matching knowledge is found in the knowledge base based on the feature vector set of the new problem, then the knowledge with the closest Euclidean distance to the feature vector set of the new problem is searched for and solved.

[0085] If no match is found in the knowledge base during the knowledge matching process, the knowledge with the closest Euclidean distance to the feature vector group of the new question is searched. This involves calculating the Euclidean distance between the feature vector group of the new question and the feature vector group of the center point of each knowledge in the knowledge base, and taking the knowledge to which the center point with the shortest Euclidean distance belongs as the approximate matching knowledge.

[0086] Step S8 involves evaluating and updating the knowledge base using the allocation scheme obtained from the retrieved knowledge, including:

[0087] Step S81: Evaluate the allocation scheme obtained by using the knowledge according to the set evaluation function, and obtain the evaluation function value of the scheme. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful.

[0088] Step S82: Update the knowledge used above using the corresponding tuples of the new problem:

[0089] The tuples of the new interception resource allocation problem are added to the subclass corresponding to the knowledge, and the knowledge corresponding to the subclass is updated. The knowledge includes feature vector range groups, adaptation rules, and centroids. Since the tuples of the new problem match the subclass, the feature vector range groups and adaptation rules of the subclass remain unchanged after the tuples of the new problem are added to the subclass. However, the centroids need to be recalculated. The calculation process is to calculate the sum of the Euclidean distances of the feature vector groups of each tuple to the feature vector groups of all other tuples, and take the feature vector group of the tuple with the smallest sum of Euclidean distances as the centroid of the subclass.

[0090] A knowledge base construction system for solving the problem of interception resource allocation includes a data acquisition module, an initialization module, a problem-solving module, a classification and clustering module, a construction module, a monitoring module, an application module, and a maintenance module.

[0091] Data acquisition module: Configured to build a model of the interception resource allocation problem to be solved, extract the features of the interception resource allocation problem, and combine all the extracted features to form a feature vector group; change the feature values ​​in the formed feature vector group to randomly generate b different interception resource allocation problem samples;

[0092] Initialization module: configured to construct a generation rule for solving the interception resource allocation problem, and to construct an evaluation function for the objective of the interception resource allocation problem;

[0093] Problem Solving Module: Configured to solve the problem samples in the data acquisition module by solving the a-type scheme generation rules in the initialization module to obtain a-type allocation schemes, and evaluate them according to the evaluation function. Record the scheme generation rules corresponding to the allocation schemes with evaluation values ​​higher than the threshold, and define them as the adaptation rules of the problem sample. Construct a "feature vector group - adaptation rule" tuple based on the adaptation rules of the problem sample.

[0094] The classification and clustering module is configured to perform classification and clustering operations on the "feature vector set - adaptation rule" pairs constructed by the problem-solving module. The classification and clustering operations are as follows: First, classify according to the adaptation rule, classify pairs with the same adaptation rule into major categories. Within each major category, perform clustering operation according to the feature vector set, and cluster pairs with similar feature vector sets into minor categories. The feature vector range set of the minor category is taken as the range of each feature vector in the minor category, and the cluster center point of the minor category is calculated.

[0095] The building module is configured to form each subclass as a knowledge item and use all the knowledge to build a knowledge base. The knowledge is in the form of "feature vector range group - adaptation rule - center point".

[0096] Listening module: Configured to listen for new intercepted resource allocation issues; if a new issue arises, trigger the application module.

[0097] Application module: configured to extract the feature vector set of the new interception resource allocation problem, input it into the knowledge base for retrieval, and use the retrieved knowledge to solve the problem and obtain the allocation scheme of the problem;

[0098] Maintenance module: Configured to evaluate the allocation scheme obtained by solving the knowledge retrieved by the application module. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful. The knowledge to which it belongs is updated according to the tuple of the new problem, and the knowledge base is updated and improved, triggering the monitoring module. If the evaluation value is lower than the set evaluation threshold, the knowledge application fails, triggering the problem solving module.

[0099] The present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it implements the knowledge base construction method for solving the interception resource allocation problem as described above.

[0100] The specific embodiments described above only illustrate the design principles of the present invention. The shapes and names of the components in this description may differ and are not limited. Therefore, those skilled in the art can modify or make equivalent substitutions to the technical solutions described in the foregoing embodiments; and these modifications and substitutions do not depart from the inventive spirit and technical solutions of the present invention, and should all fall within the protection scope of the present invention.

Claims

1. A method for constructing a knowledge base to solve the interception resource allocation problem, characterized in that... The steps of this method include: Step S1: Extract features of the interception resource allocation problem, combine all extracted features to form a feature vector group, and randomly generate each feature value in the formed feature vector group. b A different sample of interception resource allocation issues; Step S2, construct a solution to the interception resource allocation problem. a The generation rules, the constructed a All of these generation rules can generate allocation schemes to intercept resource allocation problems; Step S3: Construct an evaluation function for the target of the interception resource allocation problem, and evaluate the allocation scheme for the interception resource allocation problem generated in step S2 based on the constructed target evaluation function; Step S4: Solve the problem sample using the generation rule a constructed in step S2, obtain the allocation scheme and evaluate the allocation scheme. Define the generation rule corresponding to the allocation scheme whose evaluation function value is higher than the set threshold as the adaptation rule and construct the "feature vector group - adaptation rule" pair. Step S5: Perform classification and clustering operations on the "feature vector group - adaptation rule" binary tuples constructed in step S4; Step S6: Each subclass of clustering in step S5 is formed into a knowledge item, and all knowledge items are used to build a knowledge base; In step S5, the classification and clustering operations are as follows: First, classify according to the adaptation rules, classify the pairs with the same adaptation rules into major categories, and within each major category, perform clustering operations according to the feature vector groups, and cluster the pairs with similar feature vector groups into minor categories. The feature vector range group of the minor category is taken as the range of each feature vector in the minor category, and the cluster center point of the minor category is calculated. In step S5, binary pairs with similar feature vector groups in the major category are clustered into a minor category, and the centroid of the minor category is calculated, including: Step S51: Use the K-means algorithm to cluster the binary groups in the major class based on the feature vector set: iteratively find... k This clustering makes it possible to use this k The overall error is minimized when the mean of each cluster is used to represent the sample of the corresponding subclass. Among these, the number of clusters... k Selected based on the capacity of the knowledge base; Step S52: After the clustering operation, multiple subclasses are obtained. The vector values ​​of each feature vector group in each subclass are extracted, and the distribution range of each vector in the subclass is calculated. Thus, the value range of each feature in the feature vector group of this subclass is obtained, which serves as the criterion for judging whether a new interception resource allocation problem belongs to this subclass. Step S53: For each pair of tuples in the subclass i Calculate the pair respectively i The sum of the Euclidean distances between the feature vector set of the eigenvalue and the feature vector sets of all other binary pairs is used as the center point of the subclass. In step S6, the knowledge takes the form of: "feature vector range group - adaptation rule - center point"; In step S6, constructing the knowledge base includes: taking each subclass obtained from the above classification and clustering operations as a knowledge entry, with the knowledge entry in the form of "feature vector range group - adaptation rule - centroid". Here, the "feature vector range group" in each knowledge entry is the problem feature range of that subclass, that is, the range group of each feature in that class; the "adaptation rule" in each knowledge entry is the adaptation rule of that subclass; and the "centroid" in each knowledge entry is the centroid of that subclass. All the knowledge entries represented by the subclasses together constitute the knowledge base for the interception resource allocation problem.

2. The knowledge base construction method for solving the interception resource allocation problem according to claim 1, characterized in that: In step S1, the method for obtaining the problem sample is as follows: based on the feature vector group of the intercepted resource allocation problem, set the value range of each vector in the feature vector group, and randomly generate a feature vector group within the set value range. Each newly generated feature vector group represents a newly generated problem sample.

3. The knowledge base construction method for solving the interception resource allocation problem according to claim 1, characterized in that: In step S3, when evaluating the allocation scheme, a threshold for the evaluation function is set. When the evaluation function value is greater than the set threshold, the allocation scheme is a high-quality scheme, and the generation rule corresponding to the allocation scheme is applicable to solving the interception resource allocation problem. When the evaluation function value is not greater than the set threshold, the allocation scheme is a low-quality scheme, and the generation rule corresponding to the allocation scheme is not applicable to solving the interception resource allocation problem.

4. The knowledge base construction method for solving the interception resource allocation problem according to claim 1, characterized in that: In step S4, if a knowledge base has not yet been built, the generation rules constructed in step S2 are used to solve each problem sample generated in step S1, resulting in a allocation schemes. The a allocation schemes are evaluated according to the evaluation function constructed in step S3. The generation rules corresponding to the allocation schemes with evaluation function values ​​higher than a set threshold are recorded and defined as the adaptation rules for this problem sample. Based on the adaptation rules for each problem sample, a "feature vector group - adaptation rule" tuple is constructed. If a knowledge base has already been built, the generation rules constructed in step S2 are used to solve the newly arrived interception resource allocation problem. The a allocation schemes are evaluated according to the evaluation function constructed in step S3. The generation rules corresponding to the allocation schemes with evaluation function values ​​higher than a set threshold are recorded and defined as the adaptation rules for this problem sample. Based on the adaptation rules for each problem sample, a "feature vector group - adaptation rule" tuple is constructed. In step S4, there may be more than one adaptation rule for the interception resource allocation problem. If multiple generation rules result in allocation schemes with evaluation values ​​higher than the threshold, then all of these generation rules are recorded, and each generation rule is used to construct a "feature vector group - adaptation rule" with the problem. If the number of generation rules with evaluation values ​​of the solved allocation schemes higher than the threshold exceeds a certain threshold, then... r If a result is not obtained, only the one with the higher evaluation value will be considered. r The generation rules are recorded and binary pairs are constructed.

5. A method for solving the interception resource allocation problem using the knowledge base described in any one of claims 1-4, characterized in that: Step S7: If a new interception resource allocation problem arises, extract the feature vector set of the new problem and input the feature vector set of the new problem into the knowledge base for retrieval; use the retrieved knowledge to solve the problem and obtain the allocation scheme for the new problem. Step S8: Evaluate the allocation scheme obtained by solving the problem using the retrieved knowledge. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful. Update the knowledge according to the tuple of the new problem and update and improve the knowledge base. Proceed to step S7. If the evaluation value is lower than the set evaluation threshold, the knowledge application fails, and the process proceeds to step S4.

6. The method for solving the interception resource allocation problem using a knowledge base according to claim 5, characterized in that: The retrieval in step S7 includes: Step S71: If, based on the feature vector group of the new interception resource allocation problem, knowledge matching the input is found in the knowledge base, that is, each element in the feature vector group of the new problem is within the corresponding feature range of a certain knowledge, then the adaptation rule in this knowledge is used to solve the new problem; The matching process between the feature vector group of the new problem and the knowledge in the knowledge base involves comparing each vector value in the feature vector group of the new problem with the corresponding vector range in the feature range of each knowledge. If all the values ​​in the feature vector group of the new problem belong to the feature range group of a certain knowledge, the matching is successful; otherwise, the matching fails, and the process proceeds to step S72. In the knowledge matching process, if multiple pieces of knowledge match the new problem, the Euclidean distance between the feature vector group of the new problem and the feature vector group of the center point of the matched multiple pieces of knowledge is calculated, and the knowledge with the shortest Euclidean distance is taken as the final knowledge that matches it. Step S72: If no matching knowledge is found in the knowledge base based on the feature vector set of the new problem, then the knowledge with the closest Euclidean distance to the feature vector set of the new problem is searched for and solved. If no match is found in the knowledge base during the knowledge matching process, the knowledge with the closest Euclidean distance to the feature vector group of the new question is searched. This involves calculating the Euclidean distance between the feature vector group of the new question and the feature vector group of the center point of each knowledge in the knowledge base, and taking the knowledge belonging to the center point with the shortest Euclidean distance as the approximate matching knowledge.

7. A method for solving the interception resource allocation problem using a knowledge base according to claim 5, characterized in that: In step S8, the allocation scheme obtained by solving using the retrieved knowledge is evaluated and the knowledge base is updated and improved, including: Step S81: Evaluate the allocation scheme obtained by using the knowledge according to the set evaluation function, and obtain the evaluation function value of the scheme. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful. Step S82: Update the knowledge used above using the corresponding tuples of the new problem: The tuples of the new interception resource allocation problem are added to the subclass corresponding to the knowledge, and the knowledge corresponding to the subclass is updated. The knowledge includes feature vector range groups, adaptation rules, and centroids. Since the tuples of the new problem match the subclass, the feature vector range groups and adaptation rules of the subclass remain unchanged after the tuples of the new problem are added to the subclass. However, the centroids need to be recalculated. The calculation process is to calculate the sum of the Euclidean distances of the feature vector groups of each tuple to the feature vector groups of all other tuples, and take the feature vector group of the tuple with the smallest sum of Euclidean distances as the centroid of the subclass.

8. A knowledge base construction system for solving the interception resource allocation problem using the method of claim 1, characterized in that: The knowledge base construction system includes a data acquisition module, an initialization module, a problem-solving module, a classification and clustering module, a construction module, a monitoring module, an application module, and a maintenance module. Data acquisition module: Configured to build a model of the interception resource allocation problem, extract features of the interception resource allocation problem, and combine all extracted features to form a feature vector group; change the feature values ​​in the formed feature vector group to generate random features. b A different sample of interception resource allocation issues; Initialization module: configured to solve the interception resource allocation problem. a We will generate a set of rules and construct an evaluation function for intercepting resource allocation problems. Problem Solving Module: Configured to solve the problem samples in the data acquisition module by solving the a-type scheme generation rules in the initialization module to obtain a-type allocation schemes, and evaluate them according to the evaluation function. Record the scheme generation rules corresponding to the allocation schemes with evaluation values ​​higher than the threshold, and define them as the adaptation rules of the problem sample. Construct a "feature vector group - adaptation rule" tuple based on the adaptation rules of the problem sample. The classification and clustering module is configured to perform classification and clustering operations on the "feature vector set - adaptation rule" pairs constructed by the problem-solving module. The classification and clustering operations are as follows: First, classify according to the adaptation rule, classify pairs with the same adaptation rule into major categories. Within each major category, perform clustering operation according to the feature vector set, and cluster pairs with similar feature vector sets into minor categories. The feature vector range set of the minor category is taken as the range of each feature vector in the minor category, and the cluster center point of the minor category is calculated. The building module is configured to form each subclass as a knowledge item and use all the knowledge to build a knowledge base. The knowledge is in the form of "feature vector range group - adaptation rule - center point". Listening module: Configured to listen for new intercepted resource allocation issues; if a new issue arises, trigger the application module. Application module: configured to extract the feature vector set of the new interception resource allocation problem, input it into the knowledge base for retrieval, and use the retrieved knowledge to solve the problem and obtain the allocation scheme of the problem; Maintenance module: Configured to evaluate the allocation scheme obtained by solving the knowledge retrieved by the application module. If the evaluation value is higher than the set evaluation threshold, the knowledge application is successful. The knowledge to which it belongs is updated according to the tuple of the new problem, and the knowledge base is updated and improved, triggering the listening module. If the evaluation value is lower than the set evaluation threshold, the knowledge application fails and the problem-solving module is triggered.

Images