An intelligent review system and computer equipment for complex project clusters

Abstract

The present invention provides an intelligent review system and computer device for complex project clusters. Its distribution balancing module allocates expert nodes and projects based on a project allocation model. This project allocation model, based on a first objective function, a second objective function, and a third objective function, can achieve a three-level balance: balanced distribution of project intersections, balanced distribution of project quantities, and balanced distribution of expert balance index values. Its score optimization module, based on a scoring optimization model, optimizes expert scores by combining the normality of expert scores with expert preference information, thereby obtaining accurate project scores. The present invention can improve the efficiency and accuracy of reviews.

Description

Technical Field

[0001] The present invention belongs to the field of software information technology, and in particular relates to an intelligent review system and computer equipment for complex project clusters. Background Art

[0002] With technological advancements and policy support, the scale of scientific research project competitions is constantly expanding. To improve the effectiveness and accuracy of scientific research project reviews, it is necessary to gather experts from all regions to conduct reviews. However, requiring all experts to conduct on-site reviews would be extremely costly and difficult to implement. To address this, a distributed review system has emerged. Experts no longer need to be physically present on-site. Instead, they simply connect to the distributed network, receive the projects to be reviewed, and assign scores to complete the review, thereby improving the effectiveness of scientific research project reviews.

[0003] However, traditional review systems generally adopt project allocation methods based on keyword matching and static rules, and rely on a simple weighted average algorithm in the scoring stage. This will cause highly active expert nodes to be repeatedly assigned projects, resulting in unbalanced project distribution. This will lead to uneven loads on different expert nodes, thus affecting the efficiency of the review system. At the same time, since expert preferences and credibility differences between different expert nodes are ignored, the final project scores will be inaccurate, reducing the accuracy of the review system. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an intelligent review system and computer equipment for complex project clusters to improve the efficiency and accuracy of the review.

[0005] In a first aspect, the present invention provides an intelligent review system for complex project clusters, comprising a project library storing project information and an expert library storing expert nodes, and also comprising a distribution balance module and a score optimization module;

[0006] A distribution balance module is used to extract project information from the project library and expert nodes from the expert library, and achieve a balanced distribution of expert nodes and projects based on a pre-built project distribution model to obtain a distribution result. The project distribution model includes a first objective function for ensuring a balanced distribution of project intersections, a second objective function for ensuring a balanced distribution of project numbers, and a third objective function for ensuring a balanced distribution of expert balance index values. The distribution result includes the project and the expert node that scored the project.

[0007] The score optimization module is used to obtain the expert scores of each project after obtaining the allocation results, and optimize the expert scores based on a pre-built score optimization model to obtain the final score of each project; the score optimization model includes a first optimization model for ensuring the normality of expert scores, a second optimization model for optimizing scores in combination with expert preference information, and a third optimization model for optimizing scores in combination with score credibility.

[0008] Optionally, a balanced allocation of experts and projects is achieved based on a pre-built project allocation model, and the allocation results include:

[0009] Initialize the project sequence and expert sequence according to the project information and expert nodes; the project sequence includes multiple projects, and the expert sequence includes multiple expert nodes;

[0010] Grouping multiple expert nodes to obtain at least one expert group;

[0011] Allocate the projects in the project sequence in turn to at least one expert group to obtain an expert project matrix; the expert project matrix includes a plurality of expert nodes and a project set allocated to each expert node;

[0012] The expert project matrix is ​​used as the original population, and the allocation result is obtained based on the genetic algorithm; the fitness of the population individuals in the genetic algorithm is the negative of the value of the project allocation model.

[0013] Optionally, the allocation result is obtained based on a genetic algorithm, including:

[0014] The fitness of the original population is calculated, and based on the fitness, the roulette algorithm is used to select the population for the next iteration. After a preset number of iterations, the expert project matrix with the largest fitness is selected as the allocation result.

[0015] Optionally, the expression for the project allocation model is

[0016]

[0017] in, , , is the weight, represents the first objective function, represents the second objective function, represents the third objective function, Indicates the The balance index of expert nodes, , represents the total number of expert nodes, Represents the average value of the balance index of all expert nodes, Indicates the The expert node and The maximum number of item intersections between expert nodes, , Indicates the total number of items, Indicates the Expert nodes are assigned to review projects, Indicates the Experts were assigned to review projects, Indicates the The expert node and The minimum number of item intersections between expert nodes, Indicates the The maximum workload of project review completed by expert nodes, Indicates the The minimum workload required for an expert node to complete the project review.

[0018] Optionally, the expression of the scoring optimization model is as follows:

[0019]

[0020] in, represents the first optimization model, represents the second optimization model, represents the third optimization model, , represents the weight, represents the normalization coefficient, , Indicates the The maximum value of the scoring interval of expert nodes, Indicates the The median value of the scoring interval of expert nodes, Indicates the standard score calculation score defined by the question. Indicates the The average score of expert nodes, Indicates the The expert node Expert ratings for each project, Indicates the The expert node The subjective factor in scoring each item, Indicates the Ratings of expert nodes The standard deviation of express The minimum value of express The maximum value of No. The number of expert nodes’ scores, , Indicates the total number of ratings the project receives. Indicates the The expert node The objective factors when scoring each item, , express The expected mean of express The standard deviation of Indicates the The ratings of the expert nodes, , .

[0021] Optionally, optimize the expert scores based on a pre-built score optimization model to obtain the final score for each project, including:

[0022] For each project, the expert score of the project is input into the score optimization model, and the score optimization model outputs the final score.

[0023] Optionally, the score optimization module is also used to update the item library and expert library according to the final score.

[0024] Optionally, the intelligent review system also includes an abnormal project processing module;

[0025] An abnormal project processing module is used to construct an abnormal project determination index based on expert scores and identify abnormal projects in multiple projects based on the abnormal project determination index;

[0026] The abnormal item processing module is also used to extract the scoring features of the abnormal items according to the final scores of the abnormal items, obtain the predicted scores of the abnormal items by using the BP neural network, and use the predicted scores as the final scores of the abnormal items.

[0027] Optional abnormal item determination indicators are:

[0028]

[0029] in, Indicates the The project in The scoring stage is the abnormal item determination indicator for abnormal items, Indicates that each item is in The maximum value of the final score of the scoring stage, Indicates that each item is in The average of the final scores of the scoring stages, Indicates the The project in The final score of each scoring stage, represents the range calculation, 、 Used to characterize abnormal items with certain excellence.

[0030] In a second aspect, the present invention provides a computer device comprising the above-mentioned intelligent review system.

[0031] The beneficial effects of the present invention are:

[0032] The intelligent review system for complex project clusters provided by the present invention has an allocation balancing module that allocates expert nodes and projects based on a project allocation model. The project allocation model is based on a first objective function, a second objective function, and a third objective function, and can achieve three-layer balance: balanced distribution of project intersections, balanced distribution of project quantities, and balanced distribution of expert balance index values. It can effectively improve the balance of project allocation and balance the load of each expert node, thereby effectively improving the efficiency of the review. Its score optimization module is based on a scoring optimization model, and optimizes expert scores by combining the normality of expert scores and expert preference information. It can obtain accurate project scores, thereby improving the accuracy of the review. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a schematic diagram of the structure of an intelligent review system for complex project clusters in one embodiment of the present application;

[0034] Figure 2a This is a heat map of expert project intersection in one embodiment of the present application;

[0035] Figure 2b This is a three-dimensional graph of expert project intersection in one embodiment of the present application. DETAILED DESCRIPTION

[0036] In order to improve the efficiency and accuracy of the review system, the present invention provides an intelligent review system and computer equipment for complex project clusters; its distribution balancing module allocates expert nodes and projects based on a project allocation model. The project allocation model is based on a first objective function, a second objective function and a third objective function, and can achieve three-layer balance, namely, balanced distribution of project intersections, balanced distribution of project quantities, and balanced distribution of expert balance index values. It can effectively improve the balance of project allocation and balance the load of each expert node, thereby effectively improving the efficiency of the review; its score optimization module is based on a scoring optimization model, which combines the normality of expert scores and expert preference information to optimize expert scores, and can obtain accurate project scores, thereby improving the accuracy of the review.

[0037] To facilitate understanding, the project library and expert library included in the intelligent review system in this application are first explained.

[0038] Specifically, the project library contains project information. In some embodiments of this application, the project information includes the project and project number, and the project includes but is not limited to text data, image data, and source code. For example, in the context of a scientific research project competition, the project may be a paper, experimental data, a prototype system, a patent document, etc.

[0039] The expert database consists of multiple expert nodes, each representing an expert entity with review capabilities. Each expert entity also has identifying information, including but not limited to name and number. Each expert node is connected to the system via a unified protocol. For example, in one embodiment of the present application, the expert entity is an expert or professor in a particular field; in another embodiment of the present application, the expert entity can also be an artificial intelligence model. Example 1

[0040] In order to improve the efficiency and accuracy of the review system, the intelligent review system provided by the present invention also includes a distribution balance module and a score optimization module. Figure 1 shown.

[0041] The following combination Figure 1 Describe each module.

[0042] The allocation balance module 101 is used to extract project information from the project library, extract expert nodes from the expert library, and realize balanced allocation of expert nodes and projects based on a pre-built project allocation model to obtain an allocation result.

[0043] Specifically, the project allocation model includes a first objective function for ensuring a balanced distribution of project intersections, a second objective function for ensuring a balanced distribution of project quantities, and a third objective function for ensuring a balanced distribution of expert balance index values. The allocation results include projects and expert nodes that score the projects.

[0044] A balanced allocation of experts and projects is achieved based on a pre-built project allocation model, and the process of obtaining the allocation result includes steps 1 to 4.

[0045] Step 1: Initialize the project sequence and expert sequence based on the project information and expert information.

[0046] Exemplarily, the project sequence is obtained by randomly sorting the project numbers corresponding to multiple projects in the project library; the expert sequence is obtained by randomly sorting the identity identification information.

[0047] Step 2: Group multiple expert nodes to obtain at least one expert group.

[0048] Specifically, according to the order of the expert sequence, D expert nodes are sequentially grouped together to obtain multiple expert groups. It should be noted that in some embodiments of the present application, after multiple divisions, if the number of remaining expert nodes is less than D, the remaining expert nodes are randomly assigned to other expert groups one by one.

[0049] In other embodiments of the present application, experts may be grouped using a clustering algorithm based on pre-extracted expert features, where the expert features include but are not limited to domain labels and professional labels, and the clustering algorithm includes but is not limited to the K-Means clustering algorithm.

[0050] Step 3: Allocate the projects in the project sequence to at least one expert group in turn to obtain an expert project matrix.

[0051] In an embodiment of the present application, the expert-item matrix includes a plurality of expert nodes and an item set assigned to each expert node.

[0052] For example, {expert node A|{project a, project c, project d}, expert node B|{project a, project d, project h}, expert node C|{project a, project c, project h}}.

[0053] Step 4: Use the expert project matrix as the original population and obtain the allocation result based on the genetic algorithm.

[0054] In an embodiment of the present invention, the fitness of a population individual in the genetic algorithm is the negative of the value of the item allocation model.

[0055] Specifically, the process of obtaining the allocation result based on the genetic algorithm is as follows: calculating the fitness of the original population, and using the roulette algorithm to select the population for the next round of iteration based on the fitness, and after a preset number of iterations, selecting the expert project matrix with the largest fitness as the allocation result.

[0056] In an embodiment of the present invention, the expression of the project allocation model is:

[0057]

[0058] in, , , is the weight, represents the first objective function, represents the second objective function, represents the third objective function, Indicates the The balance index of expert nodes, , represents the total number of expert nodes, Represents the average value of the balance index of all expert nodes, Indicates the The expert node and The maximum number of item intersections between expert nodes, , Indicates the total number of items, Indicates the Expert nodes are assigned to review projects, Indicates the Experts were assigned to review Items have a value of 1 or 0. If they are assigned for review, the value is 1; if they are not assigned for review, the value is 0, which represents the actual allocation situation. Indicates the The expert node and The minimum number of item intersections between expert nodes, Indicates the The maximum workload of project review completed by expert nodes, Indicates the The minimum workload required for an expert node to complete the project review.

[0059] In different embodiments, due to different configurations of expert nodes, the number of crossovers of expert nodes is also different. In some embodiments of the present invention, the number of crossovers between any two experts is at least 1: .

[0060] Correspondingly, in the embodiment of the present application, the population fitness is calculated as The fitness meets the following conditions: single value, continuous, non-negative, maximized, reasonable, consistent, with the smallest possible computational complexity and the strongest possible versatility.

[0061] In some other embodiments of the present application, model constraints may be added to the project allocation model according to the needs of the scenario, so that the project allocation model can be adapted to various application scenarios. For example, in one application scenario, the following constraints exist:

[0062] Project review constraints: Each project must be guaranteed by Expert review, if Expert nodes were assigned to review the The corresponding items The value is 1, no review is assigned. The value is 0. The expression is:

[0063]

[0064] .

[0065] Balance constraint: any two expert nodes , There should be an intersection between them. For any expert node With any expert node The number of crossovers in the review items can be determined by the decision variable Find and define the maximum value of the crossing number , minimum , and at the same time, it is necessary to ensure that the minimum value of the crossover number is at least greater than or equal to 1. The expression is as follows:

[0066]

[0067]

[0068]

[0069] .

[0070] In some embodiments of the present invention, population iteration is based on performing a crossover operation on a selected population: randomly selecting a crossover site and exchanging expert node project data corresponding to the crossover site.

[0071] In other embodiments of the present invention, population iteration may also be based on performing crossover and mutation operations on the selected population.

[0072] The score optimization module 102 is used to obtain the expert scores of each project after obtaining the allocation results, and optimize the expert scores based on the pre-built score optimization model to obtain the final score of each project.

[0073] The scoring optimization model includes a first optimization model for ensuring the normality of expert scores, a second optimization model for optimizing scores in combination with expert preference information, and a third optimization model for optimizing scores in combination with score credibility.

[0074] Specifically, the expression of the scoring optimization model is as follows:

[0075]

[0076] in, represents the first optimization model, represents the second optimization model, represents the third optimization model, , represents the weight, represents the normalization coefficient, , Indicates the The maximum value of the scoring interval of expert nodes, Indicates the The median value of the scoring interval of expert nodes, Indicates the standard score calculation score defined by the question. Indicates the The average score of expert nodes, Indicates the The expert node Expert ratings for each project, Indicates the The expert node The subjective factor in scoring each item, Indicates the Ratings of expert nodes The standard deviation of express The minimum value of express The maximum value of No. The number of expert nodes’ scores, , Indicates the total number of times the project has been rated. Indicates the The expert node The objective factors when scoring each item, , express The expected mean of express The standard deviation of Indicates the The ratings of the expert nodes, , .

[0077] The following describes the process of optimizing expert scores based on a pre-built score optimization model to obtain the final score for each project.

[0078] Specifically, for each project, the expert score of the project is input into the score optimization model, and the score optimization model outputs the final score.

[0079] In an embodiment of the present application, in order to provide historical data guidance for subsequent reviews, the score optimization module 102 is also used to update the project library and expert library according to the final score, and then dynamically obtain the preference information of the expert node. Example 2

[0080] In another embodiment of the present invention, because each expert's assessment of a project's excellence (e.g., innovation) differs, it's difficult to reach a consensus on the project's future research prospects. This results in significant discrepancies between the scores of different expert nodes for the same project. These scoring issues are influenced by subjective biases among the expert nodes, resulting in inaccurate final scores, a problem often overlooked by traditional review systems.

[0081] To address this issue, in another embodiment of the present application, based on the first embodiment, the intelligent review system further includes an abnormal item processing module. This module is used to construct an abnormal item determination index based on the expert scores, identify abnormal items among multiple items based on the abnormal item determination index, extract the scoring features of the abnormal items based on the final scores of the abnormal items, and finally use the BP neural network to obtain a predicted score for the abnormal items, and use the predicted score as the final score of the abnormal items.

[0082] Among them, the abnormal item judgment indicators are:

[0083]

[0084] in, Indicates the The project in The scoring stage is the abnormal item determination indicator for abnormal items, Indicates that each item is in The maximum value of the final score of the scoring stage, Indicates that each item is in The average of the final scores of the scoring stages, Indicates the The project in The final score of each scoring stage, represents the range calculation, 、 Used to characterize abnormal items with certain excellence.

[0085] Artificial neurons are the basic units of BP neural networks. In the embodiments of the present application, the composition formula of neurons can be expressed as: ;in, represents the output of the neuron, represents the activation function, Indicates the Standard score The weight of Indicates bias.

[0086] In order to verify the effectiveness of the allocation balance module in the intelligent review system provided by the present invention, in one embodiment of the present application, the cross-section of the expert project review of the allocation result is visualized, such as Figure 2aand Figure 2b As shown. Combined Figure 2a and Figure 2b It can be seen that the cross-number of expert project reviews is more concentrated between 3 and 4. At the same time, the expert review set distribution is balanced, and the average cross-number of expert project reviews is 3.87, reflecting the characteristics of multi-layer balance.

[0087] In summary, the intelligent review system for complex project clusters provided by the present invention has an allocation balancing module that allocates expert nodes and projects based on a project allocation model. The project allocation model is based on the first objective function, the second objective function, and the third objective function, and can achieve three-layer balance: balanced distribution of project intersections, balanced distribution of project quantities, and balanced distribution of expert balance index values. It can effectively improve the balance of project allocation, balance the load of each expert node, and thus effectively improve the efficiency of the review; its score optimization module is based on a scoring optimization model, which combines the normality of expert scores and the expert preference information to optimize the expert scores, and can obtain accurate project scores, thereby improving the accuracy of the review.

[0088] The present invention also discloses a computer device including the above-mentioned intelligent review system. The computer device has at least the same technical effects as the above-mentioned intelligent review system for complex project clusters.

[0089] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of protection of the present application is limited to these examples. In line with the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0090] The one or more embodiments of this application are intended to encompass all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of this application should be included in the scope of protection of this application.

Claims

1. An intelligent review system for complex project clusters, comprising a project database storing project information and an expert database storing expert nodes, characterized in that: It also includes a distribution balance module and a score optimization module; The distribution balance module is used to extract the project information from the project library, extract the expert nodes from the expert library, and achieve balanced distribution of expert nodes and projects based on a pre-built project distribution model to obtain a distribution result; the project distribution model includes a first objective function for ensuring a balanced distribution of project intersections, a second objective function for ensuring a balanced distribution of project quantities, and a third objective function for ensuring a balanced distribution of expert balance index values; the distribution result includes projects and expert nodes that score the projects; The score optimization module is used to obtain the expert scores of each project after obtaining the allocation results, and optimize the expert scores based on a pre-built score optimization model to obtain the final score of each project; The scoring optimization model includes a first optimization model for ensuring the normality of expert scores, a second optimization model for optimizing scores in combination with expert preference information, and a third optimization model for optimizing scores in combination with score credibility. The expression of the scoring optimization model is as follows: in, represents the first optimization model, represents the second optimization model, represents the third optimization model, , represents the weight, represents the normalization coefficient, , Indicates the The maximum value of the scoring interval of expert nodes, Indicates the The median value of the scoring interval of expert nodes, Indicates the standard score calculation score defined by the question. Indicates the The average score of expert nodes, Indicates the The expert node Expert ratings for each project, Indicates the The expert node The subjective factor in scoring each item, Indicates the Ratings of expert nodes The standard deviation of express The minimum value of express The maximum value of No. The number of expert nodes’ scores, , Indicates the total number of times the project has been rated. Indicates the The expert node The objective factors when scoring each item, , express The expected mean of express The standard deviation of Indicates the The ratings of the expert nodes, , .

2. The intelligent review system according to claim 1, characterized in that: The balanced allocation of experts and projects based on the pre-built project allocation model is achieved, and the allocation results obtained include: Initializing a project sequence and an expert sequence according to the project information and the expert node; the project sequence includes a plurality of projects, and the expert sequence includes a plurality of expert nodes; Grouping the multiple expert nodes to obtain at least one expert group; Allocating the items in the item sequence to the at least one expert group in turn to obtain an expert item matrix; the expert item matrix includes the plurality of expert nodes and a set of items allocated to each of the expert nodes; The expert project matrix is ​​used as the original population, and the allocation result is obtained based on a genetic algorithm; the fitness of the population individuals in the genetic algorithm is the negative of the value of the project allocation model.

3. The intelligent review system according to claim 2, characterized in that: The obtaining of the allocation result by solving the problem based on a genetic algorithm includes: The fitness of the original population is calculated, and based on the fitness, a roulette wheel algorithm is used to select the population for the next iteration, and after a preset number of iterations, the expert project matrix with the largest fitness is selected as the allocation result.

4. The intelligent review system according to claim 3, characterized in that: The expression of the project allocation model is: in, , , is the weight, represents the first objective function, represents the second objective function, represents the third objective function, Indicates the The balance index of expert nodes, , represents the total number of expert nodes, Represents the average value of the balance index of all expert nodes, Indicates the The expert node and The maximum number of item intersections between expert nodes, , Indicates the total number of items, Indicates the Expert nodes are assigned to review projects, Indicates the Experts were assigned to review projects, Indicates the The expert node and The minimum number of item intersections between expert nodes, Indicates the The maximum workload of project review completed by expert nodes, Indicates the The minimum workload required for an expert node to complete the project review.

5. The intelligent review system according to claim 4, characterized in that: The expert scores are optimized based on the pre-built score optimization model to obtain the final score for each project, including: For each of the items, the expert score of the item is input into the score optimization model, and the score optimization model outputs the final score.

6. The intelligent review system according to claim 5, characterized in that: The score optimization module is further configured to update the item library and the expert library according to the final score.

7. The intelligent review system according to claim 6, characterized in that: The intelligent review system also includes an abnormal project processing module; The abnormal project processing module is configured to construct an abnormal project determination index according to the expert scores, and identify abnormal projects among the multiple projects based on the abnormal project determination index; The abnormal item processing module is further used to extract the scoring features of the abnormal item according to the final score of the abnormal item, obtain the predicted score of the abnormal item by using the BP neural network, and use the predicted score as the final score of the abnormal item.

8. The intelligent review system according to claim 7, characterized in that: The abnormal item determination indicators are: in, Indicates the The project in The scoring stage is the abnormal item determination indicator for abnormal items, Indicates that each item is in The maximum value of the final score of the scoring stage, Indicates that each item is in The average of the final scores of the scoring stages, Indicates the The project in The final score of each scoring stage, represents the range calculation, 、 Used to characterize abnormal items with certain excellence.

9. A computer device, characterized in that: The invention comprises an intelligent review system as described in any one of claims 1 to 8.

Images