Gaussian mixture target evaluation method facing intelligent city combat

Abstract

The invention relates to a Gaussian mixture target evaluation method facing intelligent city combat, and the method comprises the steps: constructing a corresponding secondary target evaluation system model according to a combat task demand, and carrying out the assignment of each sample combat target in a target sample data set according to all second-stage indexes of the secondary target evaluation system model, a second-level index assignment sequence corresponding to each sample strike target is obtained, and for each first-level index, a Gaussian mixture model is adopted to perform modeling according to a plurality of corresponding second-level indexes and the second-level index assignment sequence to obtain a corresponding first-level index model; relevant data of a to-be-evaluated strike target are substituted into the first-level index models, the evaluation score of the to-be-evaluated strike target in each first-level index is obtained, and finally calculation is carried out based on the relation between the first-level indexes and each evaluation score to obtain a final evaluation score. According to the method, the complexity of an algorithm is effectively reduced, and meanwhile, the accuracy of an evaluation result is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a Gaussian mixture target evaluation method for smart city operations. Background Art

[0002] With the rapid development of technologies such as artificial intelligence, unmanned equipment, and computer vision, modern warfare is increasingly showing new characteristics of intelligence and unmanned operations. As a place where resources are concentrated, cities have become future battlefields that countries around the world attach great importance to. How to evaluate and rank the targets based on different weapon information of the enemy and our side and factors such as the actual urban environment is the first problem to be solved in order to improve the combat effectiveness of smart cities.

[0003] At present, the solution to this problem is to quantify the threat based on the experience of military experts and a large amount of experimental data, and then establish an effective evaluation model and use a feasible method to study it. Commonly used evaluation methods include expert evaluation method and AHP method, which are not accurate and intelligent, and cannot meet the requirements of smart city operations. Summary of the invention

[0004] Based on this, it is necessary to provide a Gaussian mixture target evaluation method for smart city operations that can reduce the computational complexity while ensuring its evaluation accuracy in response to the above technical problems.

[0005] A Gaussian mixture target evaluation method for smart city operations, the method comprising:

[0006] Acquire strike mission requirements and a target sample data set, wherein the target sample data set includes a plurality of sample strike targets in the city;

[0007] A corresponding secondary target evaluation system model is constructed according to the strike mission requirements, wherein the secondary target evaluation system includes a plurality of first-level indicators, and the relationship between each first-level indicator and other first-level indicators, and a plurality of second-level indicators related to the corresponding first-level indicator are set under each first-level indicator;

[0008] Assigning a value to each sample strike target in the target sample data set according to all the second-level indicators of the second-level target evaluation system model to obtain a second-level indicator assignment sequence corresponding to each sample strike target;

[0009] For each first-level indicator, a Gaussian mixture model is used to model the corresponding multiple second-level indicators and the second-level indicator value assignment sequence to obtain a corresponding first-level indicator model;

[0010] Obtain relevant data of the target to be evaluated, and bring the relevant data into each of the first-level indicator models to obtain an evaluation score of the target to be evaluated in each first-level indicator;

[0011] A calculation is performed based on the relationship between the first-level indicators and the evaluation scores to obtain a final evaluation score of the target to be evaluated.

[0012] In one embodiment, for each first-level indicator, a Gaussian mixture model is used to model the corresponding multiple second-level indicators and the second-level indicator value sequence, including:

[0013] Constructing a training data set based on all second-level indicators under the first-level indicators and relevant data in the second-level indicator assignment sequence;

[0014] Constructing a Gaussian mixture model with multiple parameters to be solved, and iteratively solving the Gaussian mixture model according to the training data set to obtain multiple parameter value results to be solved;

[0015] According to the plurality of parameter value results and the Gaussian mixture model, a corresponding first-level indicator model is obtained.

[0016] In one embodiment, when constructing a training data set based on all second-level indicators under the first-level indicators and related data in the second-level indicator value sequence:

[0017] Determine the initial weight of each of the second-level indicators based on prior knowledge;

[0018] Extract multiple original sample assignments corresponding to the same second-level indicator from all second-level indicator assignment sequences;

[0019] Generate new sample assignments according to the original sample assignments and the initial weights of the corresponding second-level indicators, and construct a valued data set corresponding to the second-level indicators according to the original sample assignments and the new sample assignments;

[0020] The training data set is constructed according to the valued data sets of all second-level indicators corresponding to one first-level indicator.

[0021] In one embodiment, generating a new sample value according to each of the original sample values ​​and the initial weight of the corresponding second-level indicator includes:

[0022] Determine the number of times the new sample value is generated according to the initial weight of the second-level indicator;

[0023] Based on the original sample value, a random number satisfying the Gaussian distribution is added to generate the new sample value, thereby completing the generation of a new sample value;

[0024] According to the number of generation times, the generation of all new sample assignments is completed.

[0025] In one embodiment, the Gaussian mixture model with multiple parameters to be solved is expressed as:

[0026]

[0027] In the above formula, k∈[1,K] represents the component number, π k represents the weight of the kth Gaussian component, μ k represents the mean of the kth Gaussian component, Σ k represents the covariance of the kth Gaussian component, Θ=(π1,…,π K ,μ1,…,μ K ,Σ1,…,Σ K ) represents the parameter space to be solved.

[0028] In one embodiment, the relationship between each first-level indicator and other first-level indicators includes no correlation, positive correlation and negative correlation.

[0029] In one embodiment, the final evaluation score of the target to be evaluated is obtained by calculating based on the relationship between the first-level indicators and the evaluation scores, including:

[0030] According to the relationship between each of the first-level indicators and other first-level indicators, a corresponding relationship set is constructed, wherein the relationship set includes: a set of all other first-level indicators that are unrelated to the first-level indicator, a set of all other first-level indicators that are positively correlated to the first-level indicator, and a set of all other first-level indicators that are inversely correlated to the first-level indicator;

[0031] For each first-level indicator, a calculation is performed based on the corresponding relationship set and the evaluation score to obtain a corresponding benchmark score;

[0032] An average is calculated based on the benchmark scores of each of the first-level indicators, and the calculated result is used as the final evaluation score of the target to be evaluated.

[0033] In one embodiment, for each first-level indicator, the corresponding benchmark score is calculated according to the corresponding relationship set and the evaluation score, and includes:

[0034] The evaluation scores are arranged according to the relationship set and calculated using the following formula to obtain the corresponding benchmark scores:

[0035]

[0036] In the above formula, fA1 represents the benchmark score corresponding to a first-level indicator. The evaluation scores of all other first-level indicators that are irrelevant to the first-level indicators are added together. The evaluation scores of all other first-level indicators that have a positive correlation with the first-level indicator are added together. The evaluation scores corresponding to all other first-level indicators that are inversely correlated with the first-level indicators are added together, and λ1 and λ2 are pre-set adjustment parameters.

[0037] The Gaussian mixture target evaluation method for smart city operations described above constructs a corresponding secondary target evaluation system model according to the strike mission requirements, assigns a value to each sample strike target in the target sample data set according to all the second-level indicators of the secondary target evaluation system model, and obtains a secondary indicator assignment sequence corresponding to each sample strike target. For each first-level indicator, a Gaussian mixture model is used to model the corresponding multiple second-level indicators and the secondary indicator assignment sequence, and the corresponding first-level indicator model is obtained. The relevant data of the strike target to be evaluated is brought into each first-level indicator model to obtain the evaluation score of the strike target to be evaluated in each first-level indicator. Finally, the relationship between each first-level indicator and each evaluation score is calculated to obtain the final evaluation score. This method effectively reduces the complexity of the algorithm and improves the accuracy of the evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram of a process flow of a Gaussian mixture target evaluation method for smart city operations in one embodiment;

[0039] Figure 2 A schematic diagram of the structure of a secondary target evaluation system model in an embodiment;

[0040] Figure 3 A schematic diagram of a two-level target assessment system model constructed in response to a strike mission requirement in another embodiment. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0042] In view of the problem that the commonly used evaluation methods in the prior art include expert evaluation method and AHP method, which are not accurate and intelligent and cannot meet the requirements of smart city operations, in this application, Figure 1 As shown, a Gaussian mixture target evaluation method for smart city operations is provided, including the following steps:

[0043] Step S100, obtaining strike mission requirements and a target sample data set, where the target sample data set includes a plurality of sample strike targets in the city.

[0044] Step S110, constructing a corresponding secondary target evaluation system model according to the strike mission requirements, the secondary target evaluation system includes multiple first-level indicators, and the relationship between each first-level indicator and other first-level indicators, and under each first-level indicator, multiple second-level indicators related to the corresponding first-level indicator are set.

[0045] Step S120, assigning a value to each sample strike target in the target sample data set according to all second-level indicators of the second-level target evaluation system model, and obtaining a second-level indicator assignment sequence corresponding to each sample strike target.

[0046] Step S130, for each first-level indicator, a Gaussian mixture model is used to model the corresponding multiple second-level indicators and the second-level indicator value sequence to obtain a corresponding first-level indicator model.

[0047] Step S140, obtaining relevant data of the target to be evaluated, bringing the relevant data into each first-level indicator model, and obtaining the evaluation score of the target to be evaluated in each first-level indicator.

[0048] Step S150, performing calculations based on the relationships between the first-level indicators and the evaluation scores to obtain a final evaluation score of the target to be evaluated.

[0049] In step S100, the strike mission requirements and the target sample data set to be applied in the subsequent model construction are first obtained, and the sample data set includes multiple strike targets set in the city to be struck, such as a certain building, a certain square, or a certain shopping mall, etc. The strike mission requirements may include destroying certain infrastructure of the city, eliminating the city's defense forces, etc.

[0050] In step S110, a corresponding secondary target assessment system model is constructed based on the strike mission requirements. The model structure is as follows: Figure 2 As shown, it includes multiple parallel first-level indicators and multiple parallel second-level indicators under each first-level indicator.

[0051] In this embodiment, the first-level indicators can be defined from the aspects of the importance of the strike, the difficulty of the strike, the effect of the strike, etc., and the second-level indicators, for example, if the first-level indicator is the difficulty of the strike, the first-level indicator can be set to include the required weapon type, the required number of weapons, the difficulty of identification, etc. as the second-level indicators. Specifically, the secondary target evaluation system model needs to be constructed according to the requirements of each strike mission.

[0052] In one embodiment, the secondary target evaluation system model can be as follows: Figure 3 As shown, the first-level indicators include: strike importance, strike difficulty, strike effect and target characteristics. Among them, for strike importance, secondary indicators including task relevance, target matching and target obstruction are set; for strike difficulty, secondary indicators including required weapon type, required number of weapons, degree of obstruction and difficulty of identification are set; for strike effect, secondary indicators including task completion, impact on the enemy and target value are set; for target characteristics, secondary indicators including position relevance, degree of mobility and external dimensions are set.

[0053] Specifically, the first-level indicator set in the two-level target evaluation system model is recorded as {A1,…,A N}, N is the number of first-level indicators, each first-level indicator A n It is composed of several secondary indicators, denoted as M n A is the first level indicator n The number of secondary indicators included.

[0054] In this embodiment, the number of target samples in the target sample data set is 300. In other embodiments, the target sample data can be adaptively adjusted according to specific application scenarios. In this article, 300 target samples are taken as an example to illustrate this method.

[0055] In step S120, the target sample data set is represented as: {T1,…,T 300}, according to the meaning of all the second-level indicators in the two-level target evaluation system model, each target sample is assigned a corresponding value, that is, each sample target has a corresponding value in each second-level indicator, which is recorded as:

[0056] Specifically, for example, when a sample target is a square, there is a second-level indicator which is the external size of the strike target, and the sample target is assigned a corresponding size, such as an airport covers an area of ​​26 square kilometers.

[0057] Since the units of each sample target are different after being assigned values ​​based on the second-level indicators, for subsequent processing, the sequence after the initial assignment is normalized to obtain the second-level indicator assignment sequence.

[0058] In step S130, for each first-level indicator, a Gaussian mixture model is used to build a model according to the corresponding multiple second-level indicators and the secondary indicator assignment sequence, including: constructing a training data set based on all the second-level indicators under the first-level indicator and the relevant data in the secondary indicator assignment sequence, then constructing a Gaussian mixture model with multiple parameters to be solved, iteratively solving the Gaussian mixture model according to the training data set to obtain multiple parameter value results to be solved, and obtaining the corresponding first-level indicator model based on the multiple parameter value results and the Gaussian mixture model.

[0059] In the process of constructing the first-level indicator model, A n Take this as an example to illustrate.

[0060] In this embodiment, when constructing a training data set based on all second-level indicators under the first-level indicator and related data in the second-level indicator assignment sequence: first, determine the initial weight of each second-level indicator based on prior knowledge, extract multiple original sample assignments corresponding to the same second-level indicator from all second-level indicator assignment sequences, generate new sample assignments based on each original sample assignment and the initial weight of the corresponding second-level indicator, construct an assignment data set corresponding to the second-level indicator based on each original sample assignment and the new sample assignment, and construct the training data set based on the assignment data set of all second-level indicators corresponding to a first-level indicator.

[0061] Specifically, when determining the initial weights of each second-level indicator, it is known that A n Contains the second level indicator set First, based on prior knowledge or expert knowledge, the weight of the second-level indicator with the smallest effect is defined as 1. The weights of other second-level indicators are determined by comparing with this second-level indicator with the smallest effect. The weight is an integer ranging from [1,5]. The comparison process is divided into several scales: equally important weight is 1, slightly important weight is 2, medium important weight is 3, obviously important weight is 4, and very important weight is 5. Finally, the initial weights of all secondary indicators are obtained and recorded as

[0062] Specifically, the assigned data set corresponding to all sample targets is expressed as According to the initial weight of the second-level indicator, the sample target is assigned a new sample value at each second-level indicator. The generation process is based on one of the second-level indicators B n，1 Take this as an example to illustrate.

[0063] In the embodiment, generating a new sample value according to each original sample value and the initial weight of the corresponding second-level index includes: n，1The initial weight of the new sample value is determined, and the number of times the new sample value is generated is determined. Based on the original sample value, a random number that satisfies the Gaussian distribution is added to generate a new sample value, and the generation of a new sample value is completed. According to the number of times, the generation of all new sample values ​​is completed.

[0064] Specifically, the second level indicator B n，1 The initial weight is h n，1 , there are 300 sample values ​​{C n,1 (t)} t∈[1,300] , assign each new sample a value by adding a random number C n,1 ′(t)=C n,1 (t)+φ, we can get a new sample assignment, where φ is a random number that satisfies the Gaussian distribution, and perform (h n，1 -1) times of this operation, you can get a new (h n，1 -1)×300 sample values, plus the original sample value, the second level value B n，1 The corresponding new sample assignment h is obtained n，1 ×300, recorded as According to the above method, each second-level indicator will get a different number of new sample assignments. Finally, the new sample assignments of all second-level indicators form a training sample set, denoted as X = {x d} d∈[1,D] , where D represents the total number of samples;

[0065] Next, a Gaussian mixture model with multiple parameters to be solved is constructed as follows:

[0066]

[0067] In formula (1), k∈[1,K] represents the component number, π k represents the weight of the kth Gaussian component, μ k represents the mean of the kth Gaussian component, Σ k represents the covariance of the kth Gaussian component, Θ=(π1,…,π K ,μ1,…,μ K ,Σ1,…,Σ K ) represents the parameter space to be solved.

[0068] In this embodiment, when solving formula (1), if the classical EM algorithm is used, the initial value of the number of components K is M. n , the number of components will not change, but the selection of this initial value may not be accurate. Therefore, in this embodiment, an improved EM algorithm is used to solve formula (1), and the number of components can be adjusted in real time. The specific steps are as follows:

[0069] Let the initial value of the number of components be K and M n , let the weight of each component be π k The initial value is Let the mean μ of each component be k The initial value is X, which is a randomly selected sample value in the training sample set. Let the covariance of each component Σ k The initial value is Gaussian moment.

[0070] Next, using the parameter space Θ = (π1,…,π K ,μ1,…,μ K ,Σ1,…,Σ K ) and calculate the current values ​​of all parameters for each sample x d The posterior probability of each component, taking the kth Gaussian component as an example, is as follows:

[0071]

[0072] Furthermore, the parameter space Θ=(π1,…,π K ,μ1,…,μ K ,Σ1,…,Σ K ) The formula for all parameter values ​​is as follows:

[0073]

[0074] In formula (3), Represents the transpose of a matrix.

[0075] Next, calculate the distribution index Ind, the formula is as follows:

[0076]

[0077] In formula (4), sum represents the summation operator, α represents a predefined parameter. If the distribution index Ind>α×D, it means that the number of existing components is unreasonable and the number of components needs to be updated. new =K+1, select the maximum weight π k The corresponding Gaussian component is recorded as k0, and its two Gaussian components are recorded as k1 and k2. The splitting formula is as follows:

[0078]

[0079] In formula (5), ε1 and ε2 are two random numbers.

[0080] If the distribution index Ind≤α×D, it means that the current number of components is reasonable and no operation is required.

[0081] According to the above formula (2) to formula (4), an iterative process is formed. According to the preset number of times Q, this iterative process is repeated Q times to obtain the solution result of the parameter to be solved in formula (1), which is recorded as: GMM n ={K′,π1′,…,π′ K ,μ1′,…,μ′ K ,Σ′1,…,Σ′ K}.

[0082] Furthermore, the solved parameters are substituted into formula (1) to obtain the corresponding first-level indicator model, and the first-level indicators are modeled according to the above steps. Then, N different Gaussian mixture models, i.e., first-level indicator models, can be obtained, which are denoted as {GMM1,…,GMM N}.

[0083] In step S140, after obtaining the target to be evaluated and the related information, i.e., the specific second-quarter index assignment, all the related information is brought into each first-level index model, and the evaluation of each first-level index can be obtained, which is recorded as {g1,…,g N}.

[0084] In this embodiment, the relationship between each first-level indicator and other first-level indicators includes no correlation, positive correlation and negative correlation.

[0085] In step S150, calculations are performed based on the relationships between the first-level indicators and the evaluation scores to obtain the final evaluation score of the target to be evaluated, including: constructing a corresponding relationship set based on the relationships between the first-level indicators and other first-level indicators, the relationship set including: all other first-level indicator sets that are uncorrelated with the first-level indicators, all other first-level indicator sets that are positively correlated with the first-level indicators, and all other first-level indicator sets that are inversely correlated with the first-level indicators; for each first-level indicator, calculations are performed based on the corresponding relationship set and the evaluation score to obtain the corresponding benchmark score; the average is calculated based on the benchmark scores of the first-level indicators, and the calculation result is used as the final evaluation score of the target to be evaluated.

[0086] Specifically, for each first-level indicator A n , the relationship set can be represented by {A n ,J1(A n ),J2(A n ),J3(A n )} where J1(A n ) indicates that n Other unrelated first-level indicators, J2(A n ) indicates that nOther first-level indicators that are positively correlated, J3(A n ) indicates that n A collection of other first-level indicators that are inversely related.

[0087] Furthermore, taking the first-level indicator A1 as an example, the relationship set with the first-level indicator A1 is expressed as {A1, J1(A1), J2(A1), J3(A1)}, and the corresponding evaluation scores are arranged according to the relationship set to obtain: {g1, {g′2, …, g′ c1},{g c1+1 ,…,g′ c2},{gc′ 2+1 ,…,g′ N}}, and then use the following formula to calculate and get the corresponding benchmark score:

[0088]

[0089] In formula (6), fA1 represents the benchmark score corresponding to a first-level indicator. The evaluation scores of all other first-level indicators that are irrelevant to the first-level indicators are added together. The evaluation scores of all other first-level indicators that have a positive correlation with the first-level indicator are added together. The evaluation scores corresponding to all other first-level indicators that are inversely correlated with the first-level indicators are added together, and λ1 and λ2 are pre-set adjustment parameters.

[0090] Finally, after obtaining the benchmark scores of each first-level indicator {fA2,…,fA N}, the following formula is used to obtain the final evaluation score F:

[0091]

[0092] In the above-mentioned Gaussian mixture target evaluation method for smart city operations, a corresponding secondary target evaluation system model is constructed according to the strike mission requirements, and each sample strike target in the target sample data set is assigned a value according to all the second-level indicators of the secondary target evaluation system model to obtain a secondary indicator assignment sequence corresponding to each sample strike target. For each first-level indicator, a Gaussian mixture model is used to model according to the corresponding multiple second-level indicators and the secondary indicator assignment sequence to obtain the corresponding first-level indicator model, and the relevant data of the strike target to be evaluated is brought into each first-level indicator model to obtain the evaluation score of the strike target to be evaluated in each first-level indicator. Finally, based on the relationship between each first-level indicator and each evaluation score, the calculation is performed to obtain the final evaluation score. The Gaussian mixture target evaluation method proposed by this method uses a Gaussian mixture model to model the first-level indicator internally, and is divided into three different relationships externally for separate processing, which greatly reduces the complexity of the algorithm and also makes the evaluation result more accurate. It has good practical value in the field of modern smart city operations.

[0093] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0094] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0095] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0096] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A Gaussian mixture target evaluation method for smart city operations, characterized in that: The method comprises: Acquire strike mission requirements and a target sample data set, wherein the target sample data set includes a plurality of sample strike targets in the city; A corresponding secondary target evaluation system model is constructed according to the strike mission requirements, wherein the secondary target evaluation system includes a plurality of first-level indicators, and the relationship between each first-level indicator and other first-level indicators, and a plurality of second-level indicators related to the corresponding first-level indicator are set under each first-level indicator; Assigning a value to each sample strike target in the target sample data set according to all the second-level indicators of the second-level target evaluation system model to obtain a second-level indicator assignment sequence corresponding to each sample strike target; For each first-level indicator, a Gaussian mixture model is used to model the corresponding multiple second-level indicators and the second-level indicator value assignment sequence to obtain a corresponding first-level indicator model; Obtain relevant data of the target to be evaluated, and bring the relevant data into each of the first-level indicator models to obtain an evaluation score of the target to be evaluated in each first-level indicator; A calculation is performed based on the relationship between the first-level indicators and the evaluation scores to obtain a final evaluation score of the target to be evaluated.

2. The Gaussian mixture target evaluation method according to claim 1, characterized in that: For each first-level indicator, a Gaussian mixture model is used to model the corresponding multiple second-level indicators and the second-level indicator value sequence, including: Constructing a training data set based on all second-level indicators under the first-level indicators and relevant data in the second-level indicator assignment sequence; Constructing a Gaussian mixture model with multiple parameters to be solved, and iteratively solving the Gaussian mixture model according to the training data set to obtain multiple parameter value results to be solved; According to the plurality of parameter value results and the Gaussian mixture model, a corresponding first-level indicator model is obtained.

3. The Gaussian mixture target evaluation method according to claim 2, characterized in that: When constructing a training data set based on all second-level indicators under the first-level indicators and the relevant data in the second-level indicator assignment sequence: Determine the initial weight of each of the second-level indicators based on prior knowledge; Extract multiple original sample assignments corresponding to the same second-level indicator from all second-level indicator assignment sequences; Generate new sample assignments according to the original sample assignments and the initial weights of the corresponding second-level indicators, and construct a valued data set corresponding to the second-level indicators according to the original sample assignments and the new sample assignments; The training data set is constructed according to the valued data sets of all second-level indicators corresponding to one first-level indicator.

4. The Gaussian mixture target evaluation method according to claim 3, characterized in that: The generating new sample value assignment according to each of the original sample values ​​and the initial weight of the corresponding second-level indicator includes: Determine the number of times the new sample value is generated according to the initial weight of the second-level indicator; Based on the original sample value, a random number satisfying the Gaussian distribution is added to generate the new sample value, thereby completing the generation of a new sample value; According to the number of generation times, the generation of all new sample assignments is completed.

5. The Gaussian mixture target evaluation method according to any one of claims 2 to 4, characterized in that: The Gaussian mixture model with multiple parameters to be solved is expressed as: In the above formula, k∈[1,K] represents the component number, π k represents the weight of the kth Gaussian component, μ k represents the mean of the kth Gaussian component, Σ k represents the covariance of the kth Gaussian component, Θ=(π1,…,π K ,μ1,…,μ K ,Σ1,…,Σ K ) represents the parameter space to be solved.

6. The Gaussian mixture target evaluation method according to claim 5, characterized in that: The relationship between each first-level indicator and other first-level indicators includes no correlation, positive correlation and negative correlation.

7. The Gaussian mixture target evaluation method according to claim 6, characterized in that: The final evaluation score of the target to be evaluated is obtained by calculating based on the relationship between the first-level indicators and the evaluation scores, including: According to the relationship between each of the first-level indicators and other first-level indicators, a corresponding relationship set is constructed, wherein the relationship set includes: a set of all other first-level indicators that are unrelated to the first-level indicator, a set of all other first-level indicators that are positively correlated to the first-level indicator, and a set of all other first-level indicators that are inversely correlated to the first-level indicator; For each first-level indicator, a calculation is performed based on the corresponding relationship set and the evaluation score to obtain a corresponding benchmark score; An average is calculated based on the benchmark scores of each of the first-level indicators, and the calculated result is used as the final evaluation score of the target to be evaluated.

8. The Gaussian mixture target evaluation method according to claim 7, characterized in that: For each first-level indicator, the corresponding benchmark score is calculated according to the corresponding relationship set and the evaluation score, including: The evaluation scores are arranged according to the relationship set and calculated using the following formula to obtain the corresponding benchmark scores: In the above formula, fA1 represents the benchmark score corresponding to a first-level indicator. The evaluation scores of all other first-level indicators that are irrelevant to the first-level indicators are added together. The evaluation scores of all other first-level indicators that have a positive correlation with the first-level indicator are added together. The evaluation scores corresponding to all other first-level indicators that are inversely correlated with the first-level indicators are added together, and λ1 and λ2 are pre-set adjustment parameters.

Images