A Method and System for Managing Putonghua Proficiency Test Takers

Abstract

The present invention belongs to the technical field of Putonghua test management. The present invention discloses a method and system for managing Putonghua proficiency test personnel, including collecting and obtaining basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data; preprocessing the obtained basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data to obtain a test personnel information feature data set, an equipment feature data set, an examination room feature data set, and an examiner feature data set; performing weighted fusion on the test personnel information feature data set, the equipment feature data set, the examination room feature data set, and the examiner feature data set to obtain a comprehensive feature data set; training and obtaining an examination room resource prediction model according to the comprehensive feature data set, and predicting the examination room resource data; constructing an examination room resource allocation database according to the examination room resource data; which helps to reasonably allocate examination rooms, examiners, and equipment resources, and ensures the maximization of resource utilization.

Description

Technical Field

[0001] The present invention relates to the technical field of Putonghua test management. More specifically, the present invention relates to a method and system for managing personnel taking the Putonghua proficiency test. Background Art

[0002] A patent with the patent publication number CN114417293A discloses a method for managing the Putonghua proficiency test, including the following steps: S1: The management host guides candidates waiting for the exam in the examination room according to the number of candidates; S2: The management host verifies the identity by reading the candidate's ID card information and face information; if the identity verification is passed, step S3 is performed; if the identity verification fails, the administrator is notified for review; S3: The management host associates a randomly assigned exam question within the exam scope with the candidate information and prints a paper preparation reminder for the candidate; S4: The candidate randomly selects an idle exam machine for identity authentication and login; the exam machine extracts the questions assigned by the management host from the question bank according to the candidate's information and starts the exam. After the exam ends, the exam machine feeds back the status information of the exam machine to the management host. The present invention solves the problems of low efficiency, high cost, and inapplicability to the Putonghua test environment in the prior art, and provides a management method and system for the examination room of the Putonghua proficiency test, which has the characteristics of high efficiency, low cost, and being able to adapt to the Putonghua test.

[0003] The existing methods and systems for managing personnel taking the Putonghua proficiency test have the following main problems:

[0004] There is a lack of simplification of the weight coefficient tuning formula. The system needs to manually adjust the weights of each feature dataset step by step, resulting in cumbersome calculation steps, increased computational volume and time cost, and a relatively slow overall iteration speed; due to the lack of an optimization process for feature weight setting, the system may not accurately reflect the actual contribution of each feature to the model performance, resulting in unbalanced resource allocation when allocating resources and unable to achieve efficient resource scheduling;

[0005] There is no construction of a Gaussian process model, and it is impossible to accurately fit the relationship between the hyperparameter configuration and the model accuracy; this may lead to unreasonable hyperparameter configuration, affecting the final performance of the model and reducing the prediction accuracy; the lack of a mechanism for gradually optimizing the hyperparameter configuration may cause the model to stagnate in performance, making it difficult to reach the optimal state, increasing the complexity and uncertainty of model tuning; the hyperparameter search process is long and inefficient, with a slow convergence speed, wasting a large amount of computational resources and time; there is no introduction of exploration, especially insufficient search efforts in the unexplored area in the initial stage, and the model may prematurely fall into a local optimal solution, missing better hyperparameter configurations; the lack of a dynamic adjustment mechanism for the balance coefficient limit formula may lead to the inability to achieve the best balance between exploring new configurations and exploiting existing configurations, thus affecting the flexibility and adaptability of the optimization process and resulting in low efficiency;

[0006] The lack of an effective task matrix and standardized processing may lead to uneven distribution of examinees to each examination room, resulting in overloading of resources in some examination rooms while some examination rooms are idle, affecting the rational utilization of overall resources; the failure to automatically find the best allocation method may lead to conflicts among examinees in terms of time or location, further affecting the smooth progress of the examination, increasing the waiting time and inconvenience of examinees; the failure to consider factors such as examination time, equipment failure rate, and examiner absenteeism rate may generate unnecessary additional costs during allocation, leading to an increase in the overall allocation cost, thus affecting the budget and resource allocation; the lack of an intelligent resource allocation mechanism may lead to the idleness or overuse of resources such as examination rooms, examiners, and equipment, increasing resource waste and management costs; the lack of methods for row and column standardization and matrix adjustment may make the calculation process cumbersome and inefficient, increasing the time and labor costs required for resource allocation and reducing the overall efficiency of the system.

[0007] In view of this, the present invention proposes a method and system for managing Mandarin proficiency test personnel to solve the above problems. Summary of the Invention

[0008] In order to overcome the above-mentioned defects of the prior art and to achieve the above object, the present invention provides the following technical solution: A method for managing Mandarin proficiency test personnel, comprising:

[0009] S1. Obtain basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data;

[0010] S2. Preprocess the obtained basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data to obtain a test personnel information feature data set, an equipment feature data set, an examination room feature data set, and an examiner feature data set;

[0011] S3. Perform weighted fusion on the test personnel information feature data set, the equipment feature data set, the examination room feature data set, and the examiner feature data set to obtain a comprehensive feature data set; train a examination room resource prediction model based on the comprehensive feature data set to predict examination room resource data;

[0012] S4. Construct an examination room resource allocation database according to the examination room resource data;

[0013] S5. Intelligently allocate the examination time and examination location of test personnel within the next n time periods according to the examination room resource allocation database.

[0014] Further, the basic information data of the test takers includes name, ID number, contact information, age, registration time, and reserved exam time; the exam equipment data includes the number of devices, device models, device status, and device maintenance records; the examination room capacity data includes the number of examination rooms, the distribution of examination room locations, the maximum capacity of the examination rooms, and the usage status of the examination rooms; the examiner resource data includes the number of examiners, examiner qualifications, the scheduled working time of the examiners, and the scheduled working locations of the examiners.

[0015] Further, the method for preprocessing the obtained basic information data of the test takers, exam equipment data, examination room capacity data, and examiner resource data to obtain the test taker information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset includes:

[0016] Using the OPTICS algorithm to identify and remove the outliers in the basic information data of the test takers, exam equipment data, examination room capacity data, and examiner resource data, and obtaining the processed test taker information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset;

[0017] Normalizing the processed test taker information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset, and converting them into a standard normal distribution to obtain the normalized test taker information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset.

[0018] Further, the method for weighted fusion of the test taker information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset to obtain a comprehensive feature dataset includes:

[0019] Fusing the normalized test taker information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset through a weighted model to obtain a comprehensive feature dataset; denoting the test taker information feature dataset as ; denoting the equipment feature dataset as ; denoting the examination room feature dataset as ; denoting the examiner feature dataset as ;

[0020] The weighted model is: ; where is the comprehensive feature dataset; is the weight coefficient of the test taker information feature dataset; is the weight coefficient of the equipment feature dataset; is the weight coefficient of the examination room feature dataset; is the weight coefficient of the examiner feature dataset;

[0021] Design the weight coefficients of each feature dataset in the weighted model to optimize the weight coefficients of each feature dataset.

[0022] Further, the method for designing the weight coefficients of each feature dataset in the weighted model to optimize the weight coefficients of each feature dataset includes:

[0023] Adjust the weight coefficients of each feature dataset in the weighted model through the weight coefficient tuning formula; the weight coefficient tuning formula is: ; where is the weight coefficient of the th feature dataset; is the number of feature types in the th feature dataset; is the total number of all feature datasets; is the number of feature types in the th feature dataset; and are the indices of the number of feature types.

[0024] Further, the training method of the examination room resource allocation model includes;

[0025] Divide the dataset into a training set, a validation set, and a test set, and construct an examination room resource allocation model; the sample set is a subset of the dataset, and each sample set includes a historical comprehensive feature dataset and the corresponding examination room resource data; the examination room resource allocation model is a gradient boosting decision tree model;

[0026] Construct a gradient boosting decision tree model and initialize it, set the hyperparameters of the number of trees and the depth; the model fits the dataset, calculates the difference between the prediction result and the true label, and obtains the residual of the model in the current iteration; use the residual as the target variable to construct a new decision tree model; add the new decision tree model to the initial gradient boosting decision tree model and weight it according to the learning rate; repeat the steps until the preset number of iterations is reached and stop training;

[0027] Use the test set to evaluate the performance of the trained gradient boosting decision tree model, and use the coefficient of determination to evaluate and calculate the difference between the prediction result and the true label;

[0028] The coefficient of determination is ; where is the true value of the th sample in the test set; is the predicted value of the th sample predicted by the model; is the mean of all true values in the test set; is the total number of samples in the test set; is the index of the sample;

[0029] According to the evaluation results, adjust the hyperparameters of the number and depth of the tree, optimize the hyperparameters of the model, and obtain a trained examination room resource allocation model; use the trained examination room resource allocation model to predict the current comprehensive feature data to obtain the corresponding examination room resource data.

[0030] Furthermore, the method for optimizing the hyperparameters of the model includes:

[0031] S71. Define the hyperparameters to be optimized by the model as , and the optimization goal of the model is to maximize the accuracy of the model;

[0032] S72. Construct a hyperparameter space, which is a multi-dimensional grid, where each dimension represents a hyperparameter; within the hyperparameter space, randomly select a set of initial hyperparameter configurations to train the model; after each training, evaluate the performance metrics of the model by calculating the accuracy, so as to establish an initial data set;

[0033] S73. Based on the initial data set , construct a Gaussian process model to fit the relationship between the hyperparameter configuration and the accuracy, output the predicted mean and standard deviation of each configuration, and form a probability distribution model between the initial hyperparameter configuration and the accuracy;

[0034] S74. In each iteration, dynamically adjust the selection of the next hyperparameter configuration through different sampling criteria. Preset the current iteration number as , and the optimal hyperparameter configuration is ;

[0035] Increase the exploration of unknown hyperparameters through the upper confidence bound criterion formula; the upper confidence bound criterion formula is: ; where is the hyperparameter configuration selected in the th iteration; is the predicted mean of the model accuracy under the hyperparameter configuration ; is the predicted standard deviation under the hyperparameter configuration; is the balance coefficient, which is used to adjust the balance between hyperparameter exploration and exploitation;

[0036] Restrict the balance coefficient for adjusting the balance between hyperparameter exploration and exploitation through the balance coefficient restriction formula; the balance coefficient restriction formula is: ; where is the balance coefficient at the th iteration and the number of hyperparameter configurations is ; is a constant for adjusting the exploration intensity; is the number of iterations; is the number of hyperparameter configurations;

[0037] S75. Refine the dissection of the high-probability region through the probability improvement criterion formula; the probability improvement criterion formula is: ; where is the current optimal accuracy; is the model accuracy function under the hyperparameter configuration ; is a probability function used to calculate the performance improvement probability of the current candidate hyperparameter configuration; is the probability that the accuracy exceeds under the hyperparameter configuration ; is the threshold of the improvement probability, used to control the improvement amplitude of the current optimal solution;

[0038] S76. Under the selected hyperparameter configuration , train the model and calculate the accuracy , obtain the performance metrics of the current configuration, add the new hyperparameter configuration and the corresponding performance metrics to the dataset, and retrain the Gaussian process model using the updated dataset ;

[0039] S77. When the iteration reaches the preset number of iterations, select the hyperparameter configuration with the best accuracy as the optimal configuration of the final model.

[0040] Furthermore, the method for constructing the examination room resource allocation database based on the examination room resource data includes:

[0041] The examination room resource allocation database is a relational database MySQL; design database tables based on the obtained examination room resource data, associate the created database tables, and insert specific examination room resource data using SQL language; create an API framework using Flask, write processing functions at each API endpoint to implement the functions of adding, deleting, modifying, and querying data, and obtain the built examination room resource allocation database.

[0042] Furthermore, the method for intelligently allocating the examination time and examination location of testers within the next n time periods based on the examination room resource allocation database includes:

[0043] S91. Construct a task matrix based on the examination room resource allocation database, and the task matrix is: ; where is the task matrix; is the number of testers; is the number of examination rooms; The cost value of assigning the tester in the th row to the th examination room;

[0044] S92. For each row of the task matrix , find the minimum value of this row through the row normalization formula, and subtract it from all elements of this row to make at least one element in each row zero; the row normalization formula is: ; where is the cost value of assigning the tester in the th row to the th examination room; is the minimum value of the th row, that is, the minimum cost value among the costs of assigning the tester in the th position to all examination rooms;

[0045] For each column of the task matrix , find the minimum value of this column through the column normalization formula, and subtract it from all elements of this row to make at least one element in each column zero; the column normalization formula is: ; where is the minimum value of the th column, that is, the minimum cost value among the assignment costs of all testers in the th examination room;

[0046] Adjust the cost value of assigning the tester in the th row to the th examination room through the dynamic cost update formula; the dynamic cost update formula is: ; where is the cost value after dynamic adjustment; is the proportionality coefficient used to adjust the scale of the overall cost; is the examination time; is the equipment failure rate of the examination room; is the examiner absenteeism rate of the examination room; is the number of examination locations;

[0047] S93. Check the distribution of zero elements in each row and each column, and use the fewest horizontal and vertical lines to cover all zero elements; define as the number of lines required for the fewest coverage. If , then find the optimal assignment plan and the algorithm terminates; if , then matrix adjustment is required; where is the and the smaller value of;

[0048] S94. Find the minimum value among the uncovered elements , and adjust the elements of the task matrix; for the uncovered elements, subtract the minimum value ; for the elements covered by two straight lines, add the minimum value , and adjust the task matrix to generate new zero elements;

[0049] S95. Repeat steps S94 - S95 until horizontal lines and vertical lines can cover all zero elements, that is ; select non - conflicting zero elements from the final matrix, with only one zero element selected for each row and each column to form an optimal resource allocation plan.

[0050] A Putonghua proficiency test personnel management system, comprising:

[0051] A data acquisition module, used to acquire basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data;

[0052] A data processing module, used to pre - process the acquired basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data to obtain a test personnel information feature data set, an equipment feature data set, an examination room feature data set, and an examiner feature data set;

[0053] An examination room resource prediction module, used to perform weighted fusion on the test personnel information feature data set, the equipment feature data set, the examination room feature data set, and the examiner feature data set to obtain a comprehensive feature data set; train and obtain an examination room resource prediction model according to the comprehensive feature data set, and predict the examination room resource data;

[0054] An examination room resource allocation database construction module, used to construct an examination room resource allocation database according to the examination room resource data;

[0055] An intelligent allocation module, used to intelligently allocate the examination time and examination location of test personnel within the next n time periods according to the examination room resource allocation database.

[0056] The technical effects and advantages of the Putonghua proficiency test personnel management method and system of the present invention:

[0057] By introducing a weight coefficient tuning formula, the process of weight adjustment is simplified, unnecessary calculation steps are reduced, the iteration speed of the model is accelerated, and the overall calculation efficiency of the resource allocation system is improved; different weights are automatically assigned according to the number of feature types, avoiding the subjectivity of manually setting weights, and making the contribution of features more scientific and reasonable;

[0058] By constructing a Gaussian process model, the relationship between hyperparameter configurations and model accuracy can be more accurately fitted; by gradually optimizing hyperparameter configurations, it is ensured that the final configuration of the model can achieve higher accuracy; by phased sampling criteria (such as the upper confidence bound criterion and the probability improvement criterion), the balance between exploration and exploitation is adjusted at different iteration stages, effectively shortening the hyperparameter search process and improving the convergence speed; by introducing exploration in the upper confidence bound criterion formula, especially increasing the search intensity for the unexplored regions of the hyperparameter space in the initial stage, it helps to discover potential better configurations and reduce the risk of local optima; using the balance coefficient constraint formula, the balance between exploration and exploitation can be dynamically adjusted, achieving the best trade-off between exploring new configurations and exploiting existing excellent configurations, thus making the optimization process more adaptable;

[0059] Through the steps of the task matrix and normalization processing, the best allocation method is automatically found to ensure that each candidate can be assigned to a suitable examination room, avoid conflicts, and improve the allocation efficiency; the dynamic cost update formula takes into account factors such as examination time, equipment failure rate, and examiner absenteeism rate, thus optimizing the cost when candidates are assigned to examination rooms and minimizing the overall allocation cost; through intelligent resource allocation, the reasonable utilization of resources such as examination rooms, examiners, and equipment is guaranteed, making the resource allocation more balanced and reducing the vacancy rate of examination rooms; adopting an automated intelligent allocation method reduces the errors that may be brought by manual allocation, ensuring the accuracy and fairness of the allocation; through row-column normalization and matrix adjustment, the allocation range is quickly narrowed, making the calculation process more concise and efficient, thus greatly shortening the resource allocation time. Brief Description of the Drawings

[0060] Figure 1 It is a schematic flowchart of a method for managing Putonghua proficiency test personnel according to the present invention;

[0061] Figure 2 It is a schematic structural diagram of a system for managing Putonghua proficiency test personnel according to the present invention. Detailed Embodiments

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0063] Embodiment 1: Please refer to Figure 1 As shown, a method for managing Putonghua proficiency test personnel in this embodiment includes:

[0064] S1. Obtain basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data;

[0065] S2. Preprocess the obtained basic information data of testers, examination equipment data, examination room capacity data, and examiner resource data to obtain a tester information feature dataset, an equipment feature dataset, an examination room feature dataset, and an examiner feature dataset;

[0066] S3. Weightedly fuse the tester information feature dataset, the equipment feature dataset, the examination room feature dataset, and the examiner feature dataset to obtain a comprehensive feature dataset; Train and obtain an examination room resource prediction model based on the comprehensive feature dataset, and predict the examination room resource data;

[0067] S4. Construct an examination room resource allocation database based on the examination room resource data;

[0068] S5. Intelligently allocate the examination time and examination location of testers within the next n time periods according to the examination room resource allocation database.

[0069] The basic information data of testers includes name, ID number, contact information, age, registration time, and reserved examination time; The examination equipment data includes the number of equipment, equipment models, equipment status, and equipment maintenance records; The examination room capacity data includes the number of examination rooms, the distribution of examination room locations, the maximum capacity of examination rooms, and the usage status of examination rooms; The examiner resource data includes the number of examiners, examiner qualifications, examiner work arrangement time, and examiner work arrangement location.

[0070] The basic information data of testers is obtained by querying through the Putonghua proficiency tester management terminal; The examination equipment data is collected through IoT sensors deployed in the examination rooms; The examination room capacity data and examiner resource data are obtained by recording in the MySQL database;

[0071] The method for preprocessing the obtained basic information data of testers, examination equipment data, examination room capacity data, and examiner resource data to obtain a tester information feature dataset, an equipment feature dataset, an examination room feature dataset, and an examiner feature dataset includes:

[0072] Use the OPTICS algorithm to identify and remove the outliers in the basic information data of testers, examination equipment data, examination room capacity data, and examiner resource data to obtain the processed tester information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset;

[0073] Normalize the processed tester information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset, and convert them into a standard normal distribution to obtain the normalized tester information feature dataset, equipment feature dataset, examination room feature dataset, and examiner feature dataset.

[0074] A method for weighted fusion of the tester information feature dataset, device feature dataset, examination room feature dataset, and examiner feature dataset to obtain a comprehensive feature dataset includes:

[0075] Fusing the normalized tester information feature dataset, device feature dataset, examination room feature dataset, and examiner feature dataset through a weighted model to obtain a comprehensive feature dataset; Denote the tester information feature dataset as , Denote the device feature dataset as , Denote the examination room feature dataset as ; Denote the examiner feature dataset as ;

[0076] The weighted model is: ; Wherein, is the comprehensive feature dataset; is the weight coefficient of the tester information feature dataset; is the weight coefficient of the device feature dataset; is the weight coefficient of the examination room feature dataset; is the weight coefficient of the examiner feature dataset;

[0077] Design the weight coefficients of each feature dataset in the weighted model to optimize the weight coefficients of each feature dataset.

[0078] A method for designing the weight coefficients of each feature dataset in the weighted model to optimize the weight coefficients of each feature dataset includes:

[0079] Adjust the weight coefficients of each feature dataset in the weighted model through the weight coefficient tuning formula; The weight coefficient tuning formula is: ; Wherein, is the weight coefficient of the th feature dataset; is the number of feature types in the th feature dataset; is the total number of all feature datasets; is the number of feature types in the th feature dataset; and are the indexes of the number of feature types;

[0080] For example, the number of feature types in the tester information feature dataset is 3, the number of feature types in the device feature dataset is 5, the number of feature types in the examination room feature dataset is 2, and the number of feature types in the examiner feature dataset is 4;

[0081] Then the weight coefficient , the weight coefficient of the device feature dataset , the weight coefficient of the examination room feature dataset , the weight coefficient of the examiner feature dataset ; Then verify that the sum of these weight coefficients is 0.21 + 0.36 + 0.14 + 0.29 = 1.

[0082] The training method of the examination room resource allocation model includes;

[0083] Divide the dataset into a training set, a validation set, and a test set, and construct an examination room resource allocation model; The sample set is a subset of the dataset, and each sample set includes a historical comprehensive feature dataset and the corresponding examination room resource data; The examination room resource allocation model is a gradient boosting decision tree model;

[0084] Construct a gradient boosting decision tree model and initialize it, and set the hyperparameters of the number and depth of the tree; The model fits the dataset, and by calculating the difference between the prediction result and the true label, the residual of the model in the current iteration is obtained; Use the residual as the target variable to construct a new decision tree model; Add the new decision tree model to the initial gradient boosting decision tree model and weight it according to the learning rate; Repeat the steps until the preset number of iterations is reached and stop training;

[0085] Use the test set to evaluate the performance of the trained gradient boosting decision tree model, and use the coefficient of determination to evaluate and calculate the difference between the prediction result and the true label;

[0086] The coefficient of determination is ; Among them, is the true value of the th sample in the test set; is the predicted value of the th sample predicted by the model; is the mean of all true values in the test set; is the total number of samples in the test set; is the index of the sample;

[0087] According to the evaluation results, adjust the hyperparameters of the number and depth of the tree, optimize the hyperparameters of the model, and obtain the trained examination room resource allocation model; Use the trained examination room resource allocation model to predict the current comprehensive feature data to obtain the corresponding examination room resource data.

[0088] The method for optimizing the hyperparameters of the model includes:

[0089] S71. Define the hyperparameters that need to be optimized by the model as , and the optimization goal of the model is to maximize the accuracy of the model;

[0090] S72. Construct a hyperparameter space, which is a multi-dimensional grid where each dimension represents a hyperparameter; randomly select a set of initial hyperparameter configurations within the hyperparameter space to train the model; after each training, evaluate the performance metrics of the model by calculating the accuracy rate, thereby establishing an initial dataset;

[0091] S73. Based on the initial dataset , construct a Gaussian process model to fit the relationship between hyperparameter configurations and accuracy rate, output the predicted mean and standard deviation for each configuration, and form a probability distribution model between the initial hyperparameter configurations and accuracy rate;

[0092] S74. In each iteration, dynamically adjust the selection of the next hyperparameter configuration through different sampling criteria. Preset the current iteration number as , and the optimal hyperparameter configuration is ;

[0093] Increase the exploration of unknown hyperparameters through the upper confidence bound criterion formula; the upper confidence bound criterion formula is: ; where is the hyperparameter configuration selected in the -th iteration; is the predicted mean of the model accuracy rate under the hyperparameter configuration ; is the predicted standard deviation under the hyperparameter configuration; is the balance coefficient used to adjust the balance between hyperparameter exploration and exploitation;

[0094] Limit the balance coefficient for adjusting the balance between hyperparameter exploration and exploitation through the balance coefficient limit formula; the balance coefficient limit formula is: ; where is the balance coefficient at the -th iteration and when the number of hyperparameter configurations is ; is a constant for adjusting the exploration intensity; is the iteration number; is the number of hyperparameter configurations;

[0095] For example, the constant for adjusting the exploration intensity is 30. At the 4th iteration, when the number of hyperparameter configurations is 10, the balance coefficient at this time is ;

[0096] S75. Improve the refinement of the dissection of the high-probability region through the probability improvement criterion formula; the probability improvement criterion formula is: ; where is the current optimal accuracy rate; is the accuracy rate under the hyperparameter configuration The model accuracy function below; Is a probability function used to calculate the probability of performance improvement for the current candidate hyperparameter configuration; Is for the hyperparameter configuration The probability that the accuracy exceeds Under; Is the threshold of the improvement probability, used to control the improvement amplitude for the current optimal solution;

[0097] The size of this value affects the exploration tendency when selecting the next candidate configuration; if Is larger, a configuration that can significantly improve the accuracy needs to be found; if it is smaller, a smaller improvement is allowed to more finely optimize the configuration near the known optimal solution;

[0098] S76. Under the selected hyperparameter configuration Train the model and calculate the accuracy , obtain the performance metrics of the current configuration, add the new hyperparameter configuration and the corresponding performance metrics to the dataset, and retrain the Gaussian process model using the updated dataset ;

[0099] S77. When the iteration reaches the preset number of iterations, select the hyperparameter configuration with the best accuracy as the optimal configuration of the final model.

[0100] The method for constructing an examination room resource allocation database based on examination room resource data includes:

[0101] The said examination room resource allocation database is a relational database MySQL; design database tables based on the obtained examination room resource data, associate the created database tables, and insert specific examination room resource data using SQL language; create an API framework with Flask, write processing functions at each API endpoint to implement the functions of adding, deleting, modifying, and querying data, and obtain the built examination room resource allocation database.

[0102] The method for intelligently allocating the examination time and examination location of testers in the next n time periods according to the examination room resource allocation database includes:

[0103] S91. Construct a task matrix according to the examination room resource allocation database, and the task matrix is: ; where Task matrix; Is the number of testers; Is the number of examination rooms; Is to allocate the tester in the Row to the th examination room;

[0104] S92. For the task matrix For each row, find the minimum value of that row through the row normalization formula and subtract it from all elements of that row, so that at least one element in each row is zero; the row normalization formula is: ; where is the cost value of the tester in the th row assigned to the th examination room; is the minimum value of the th row, that is, among the costs of the tester in the th position assigned to all examination rooms, the minimum cost value;

[0105] For each column of the task matrix , find the minimum value of that column through the column normalization formula and subtract it from all elements of that column, so that at least one element in each column is zero; the column normalization formula is: ; where is the minimum value of the th column, that is, among the allocation costs of all testers in the th examination room, the minimum cost value;

[0106] Adjust the cost value of the tester in the th row assigned to the th examination room through the dynamic cost update formula; the dynamic cost update formula is: ; where is the cost value after dynamic adjustment; is the proportionality coefficient used to adjust the scale of the overall cost; is the examination time; is the equipment failure rate of the examination room ; is the examiner absenteeism rate of the examination room ; is the number of examination locations;

[0107] As and increase, their impact on the cost will not increase linearly but gradually slow down, thus avoiding excessive influence of certain extreme values on the cost; for the number of examination locations , use the logarithmic function to make its cost reduction effect decrease, and avoid reducing the cost infinitely by increasing the number of examination rooms;

[0108] For example, if the examination time is 3 hours, the equipment failure rate of the examination room is 5%, the examiner absenteeism rate of the examination room is 10%, the number of examination locations is 3, and the proportionality coefficient is 10, then the cost value after dynamic adjustment;

[0109] S93. Examine the distribution of zero elements in each row and each column, and use the minimum number of horizontal and vertical lines to cover all zero elements; define as the number of lines required for the minimum coverage. If , then find the optimal allocation scheme and terminate the algorithm; if , then matrix adjustment is required; where is and the smaller value of;

[0110] S94. Find the minimum value among the uncovered elements, and adjust the elements of the task matrix; for the uncovered elements, subtract the minimum value ; for the elements covered by two lines, add the minimum value to adjust the task matrix to generate new zero elements;

[0111] S95. Repeat steps S94 - S95 until all zero elements can be covered by horizontal lines and vertical lines, that is ; select non - conflicting zero elements from the final matrix, with only one zero element selected from each row and each column to form the optimal resource allocation scheme.

[0112] In this embodiment, by introducing the weight coefficient tuning formula, the process of weight adjustment is simplified, unnecessary calculation steps are reduced, the iteration speed of the model is accelerated, and the overall calculation efficiency of the resource allocation system is improved; different weights are automatically assigned according to the number of feature types, avoiding the subjectivity of manually setting weights and making the contribution of features more scientific and reasonable;

[0113] By constructing a Gaussian process model, the relationship between hyperparameter configuration and model accuracy can be more accurately fitted; by gradually optimizing the hyperparameter configuration, it is ensured that the final configuration of the model can achieve higher accuracy; by the sampling criteria in different stages (such as the upper confidence bound criterion and the probability improvement criterion), the balance between exploration and exploitation is adjusted in different iteration stages, effectively shortening the hyperparameter search process and improving the convergence speed; by introducing exploration in the upper confidence bound criterion formula, especially increasing the search intensity for the unexplored areas of the hyperparameter space in the initial stage, it helps to discover potential better configurations and reduce the risk of local optimality; using the balance coefficient constraint formula, the balance between exploration and exploitation can be dynamically adjusted, achieving the best trade - off between exploring new configurations and exploiting existing excellent configurations, so that the optimization process is more adaptable;

[0114] Automatically find the best allocation method through the steps of task matrix and normalization processing, ensure that each candidate can be allocated to a suitable examination room, avoid conflicts, and improve the allocation efficiency; the dynamic cost update formula takes into account factors such as examination time, equipment failure rate, and examiner absenteeism rate, so as to optimize the cost when candidates are allocated to examination rooms and minimize the overall allocation cost; through intelligent resource allocation, ensure the reasonable utilization of resources such as examination rooms, examiners, and equipment, make the resource allocation more balanced, and reduce the vacancy rate of examination rooms; adopt an automated intelligent allocation method, reduce the errors that may be brought by manual allocation, and ensure the accuracy and fairness of the allocation; through row and column normalization and matrix adjustment, quickly narrow the allocation range, make the calculation process more concise and efficient, and thus greatly shorten the resource allocation time.

[0115] Embodiment 2: Please refer to Figure 2 As shown in the figure, a Putonghua proficiency test personnel management system in this embodiment includes:

[0116] A data acquisition module, used to acquire basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data;

[0117] A data processing module, used to preprocess the acquired basic information data of test personnel, examination equipment data, examination room capacity data, and examiner resource data to obtain a test personnel information feature data set, an equipment feature data set, an examination room feature data set, and an examiner feature data set;

[0118] An examination room resource prediction module, used to perform weighted fusion on the test personnel information feature data set, the equipment feature data set, the examination room feature data set, and the examiner feature data set to obtain a comprehensive feature data set; train and obtain an examination room resource prediction model according to the comprehensive feature data set, and predict the examination room resource data;

[0119] An examination room resource allocation database construction module, used to construct an examination room resource allocation database according to the examination room resource data;

[0120] An intelligent allocation module, used to intelligently allocate the examination time and examination location of test personnel within the next n time periods according to the examination room resource allocation database.

[0121] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for ordinary technical users in the technical field, several improvements and refinements made without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for managing people taking a Putonghua proficiency test, characterized in that: include: S1. Obtain basic information data of testers, test equipment data, test room capacity data and examiner resource data; S2, preprocessing the acquired tester basic information data, test equipment data, test room capacity data and examiner resource data to obtain a tester information feature data set, a device feature data set, a test room feature data set and an examiner feature data set; S3, weighted fusion of the tester information feature data set, the equipment feature data set, the examination room feature data set and the examiner feature data set to obtain a comprehensive feature data set; training a test room resource prediction model based on the comprehensive feature data set to predict the test room resource data; S4, constructing an examination room resource allocation database according to the examination room resource data; S5. Intelligently allocate the examination time and examination location of the testers in the next n time periods according to the examination room resource allocation database; The method of weightedly fusing the tester information feature data set, the equipment feature data set, the examination room feature data set and the examiner feature data set to obtain a comprehensive feature data set includes: The normalized tester information feature data set, device feature data set, examination room feature data set and examiner feature data set are fused through a weighted model to obtain a comprehensive feature data set; the tester information feature data set is recorded as V1, the device feature data set is recorded as V2, the examination room feature data set is recorded as V3; and the examiner feature data set is recorded as V4; The weighted model is: EQ=V1·α1+V2·α2+V3·α3+V4·α4; wherein EQ is a comprehensive feature data set; α1 is a weight coefficient of a tester information feature data set; α2 is a weight coefficient of a device feature data set; α3 is a weight coefficient of a test site feature data set; α4 is a weight coefficient of an examiner feature data set; The weight coefficient of each feature data set in the weighted model is designed to optimize the weight coefficient of each feature data set; The training method of the examination room resource prediction model includes: The data set is divided into a training set, a validation set and a test set, and an examination room resource prediction model is constructed; the sample set is a subset of the data set, and each sample set includes a historical comprehensive feature data set and corresponding examination room resource data; the examination room resource prediction model is a gradient boosting decision tree model; Build a gradient boosted decision tree model and initialize it, setting the number of trees and depth hyperparameters; fit the model to the data set, and calculate the difference between the predicted result and the true label to get the residual of the model in the current iteration; use the residual as the target variable to build a new decision tree model; add the new decision tree model to the initial gradient boosted decision tree model and weight it according to the learning rate; repeat the steps until the preset number of iterations is reached and stop training; Use the test set to evaluate the performance of the trained gradient boosting decision tree model, and use the coefficient of determination to evaluate the difference between the predicted results and the true labels; The coefficient of determination is Among them, yi ′ is the true value of the i′th sample in the test set; is the predicted value of the i′th sample predicted by the model; is the mean of all true values ​​in the test set; n is the total number of samples in the test set; i′ is the index of the sample; According to the evaluation results, the number of trees and the depth hyperparameters are adjusted to optimize the model's hyperparameters to obtain a trained exam room resource prediction model; the trained exam room resource prediction model is used to predict the current comprehensive feature data to obtain the corresponding exam room resource data; The method of designing the weight coefficient of each feature data set in the weighted model so that the weight coefficient of each feature data set reaches the optimum includes: The weight coefficient of each feature data set in the weighted model is adjusted by the weight coefficient tuning formula; the weight coefficient tuning formula is: Among them, α o is the weight coefficient of the oth feature data set; g o is the number of feature types in the oth feature data set; H is the total number of all feature data sets; g u is the number of feature types in the u-th feature data set; o and u are the indexes of the number of feature types; The method for tuning the hyperparameters of the model includes: S71. Define the hyperparameter that needs to be tuned for the model as η, and the optimization goal of the model is to maximize the accuracy of the model; S72. construct a hyperparameter space, where the hyperparameter space is a multidimensional grid, in which each dimension represents a hyperparameter; in the hyperparameter space, randomly select a set of initial hyperparameter configurations to train the model; after each training, evaluate the performance indicators of the model by calculating the accuracy, thereby establishing an initial data set; S73. Based on the initial data set B, a Gaussian process model is constructed to fit the relationship between the hyperparameter configuration and the accuracy, and the predicted mean and standard deviation of each configuration are output to form a probability distribution model between the initial hyperparameter configuration and the accuracy. S74. In each iteration, dynamically adjust and select the next hyperparameter configuration through different sampling criteria, preset the current number of iterations to T, and the optimal hyperparameter configuration to η T ; The exploration of unknown hyperparameters is enhanced by using the upper confidence bound criterion formula; the upper confidence bound criterion formula is: Among them, η T+1 is the hyperparameter configuration selected in the T+1th iteration; μ(η) is the predicted mean of the model accuracy under the hyperparameter configuration η; σ(η) is the predicted standard deviation under the hyperparameter configuration; ω is the balance coefficient, which is used to adjust the balance between hyperparameter exploration and utilization; The balance coefficient for adjusting the exploration and utilization of hyperparameters is limited by the balance coefficient limitation formula; the balance coefficient limitation formula is: Where ω(T′,Q) is the balance coefficient at the T′th iteration and the number of hyperparameter configurations is Q; V is a constant for adjusting the exploration intensity; T′ is the number of iterations; Q is the number of hyperparameter configurations; S75, improving the segmentation and refinement of the high probability area by using a probability improvement criterion formula; the probability improvement criterion formula is: Among them, f(η * ) is the current optimal accuracy; f(η) is the model accuracy function under the hyperparameter configuration η; Pr is the probability function used to calculate the performance improvement probability of the current candidate hyperparameter configuration; Pr(f(η)>f(η * )+ξ) is the accuracy that exceeds f(η) under the hyperparameter configuration η * )+ξ; ξ is the threshold of the probability of improvement, which is used to control the improvement of the current optimal solution; S76. In the selected hyperparameter configuration η T+1 Next, train the model and calculate the accuracy f(η T+1 ), obtain the performance index of the current configuration, add the new hyperparameter configuration and the corresponding performance index to the data set, and retrain the Gaussian process model using the updated data set B; S77. When the iteration reaches a preset number of iterations, the hyperparameter configuration with the best accuracy is selected as the optimal configuration of the final model; The method for intelligently allocating the examination time and examination location of testers in the future n time periods according to the examination room resource allocation database includes: S91. Construct a task matrix according to the examination room resource allocation database. The task matrix is: Among them, C is the task matrix; m is the number of testers; n is the number of examination rooms; C mn is the cost value of assigning the mth row of testers to the nth examination room; S92. For each row of the task matrix C, find the minimum value of the row by using the row normalization formula, and subtract it from all elements of the row so that at least one element in each row is zero; the row normalization formula is: ij =C ij -min(C i,： ), where C ij The cost value of assigning the i-th row of testers to the j-th examination room; min(C i,： ) is the minimum value of the i-th row, that is, the minimum cost value among the costs assigned to all examination rooms by the i-th tester; For each column of the task matrix C, find the minimum value of the column through the column normalization formula and subtract it from all elements of the row so that at least one element in each column is zero; the column normalization formula is: ij =C ij -min(C ：,j ); where min(C′ ：,j ) is the minimum value of the jth column, that is, the minimum cost value among the allocated costs of all testers in the jth examination room; The cost value C of assigning the i-th row of testers to the j-th examination room is calculated by the dynamic cost update formula. ij Adjustment is made; the dynamic cost update formula is: Among them, C i ' j is the cost value after dynamic adjustment; γ is the proportional coefficient, which is used to adjust the scale of the overall cost; t is the test time; F j is the equipment failure rate of examination room j; A j is the examiner absence rate of examination room j; D is the number of examination locations; S93. Check the distribution of zero elements in each row and column, and use the least horizontal and vertical straight lines to cover all zero elements; define k as the minimum number of straight lines required for covering, if k = min(m,n), then the optimal allocation scheme is found and the algorithm terminates; if k < min(m,n), the matrix needs to be adjusted; where min(m,n) is the smaller value of m and n; S94, find the minimum value δ among the uncovered elements, and adjust the task matrix elements; for the uncovered elements, subtract the minimum value δ; for the elements covered by the two straight lines, add the minimum value δ, and adjust the task matrix to generate new zero elements; S95. Repeat steps S94-S95 until all zero elements can be covered by m horizontal lines and n vertical lines, that is, k=min(m,n); select non-conflicting zero elements from the final matrix so that only one zero element is selected for each row and column, forming an optimal resource allocation plan.

2. A method for managing people taking a Putonghua proficiency test according to claim 1, characterized in that: The basic information data of the tester includes name, ID number, contact information, age, registration time and scheduled examination time; the examination equipment data includes the number of equipment, equipment model, equipment status and equipment maintenance records; the examination room capacity data includes the number of examination rooms, examination room location distribution, maximum examination room capacity and examination room usage status; the examiner resource data includes the number of examiners, examiner qualifications, examiner work schedule and examiner work schedule location.

3. A method for managing people taking a Putonghua proficiency test according to claim 2, characterized in that: The method of preprocessing the acquired tester basic information data, test equipment data, test room capacity data and examiner resource data to obtain a tester information feature data set, a device feature data set, a test room feature data set and an examiner feature data set includes: Use the OPTICS algorithm to identify and remove outliers in the tester basic information data, test equipment data, test room capacity data, and examiner resource data, and obtain the processed tester information feature data set, equipment feature data set, test room feature data set, and examiner feature data set; The processed tester information feature data set, device feature data set, examination room feature data set and examiner feature data set are normalized and converted into standard normal distribution to obtain normalized tester information feature data set, device feature data set, examination room feature data set and examiner feature data set.

4. A method for managing people taking a Putonghua proficiency test according to claim 3, characterized in that: The method for constructing an examination room resource allocation database according to examination room resource data comprises: The examination room resource allocation database is a relational database MySQL; a database table is designed based on the acquired examination room resource data, the created database table is associated, and the specific examination room resource data is inserted using the SQL language; Flask is used to create an API framework, and a processing function is written at each API endpoint to implement the data addition, deletion, modification and query functions, thereby obtaining a constructed examination room resource allocation database.

5. A management system for personnel of Mandarin proficiency test, used to implement a management method for personnel of Mandarin proficiency test as claimed in any one of claims 1 to 4, characterized in that: include: The data acquisition module is used to obtain the basic information data of the testers, the test equipment data, the test room capacity data and the examiner resource data; The data processing module is used to pre-process the acquired tester basic information data, test equipment data, test room capacity data and examiner resource data to obtain a tester information feature data set, a device feature data set, a test room feature data set and an examiner feature data set; The examination room resource prediction module is used to perform weighted fusion on the tester information feature data set, the equipment feature data set, the examination room feature data set and the examiner feature data set to obtain a comprehensive feature data set; train the examination room resource prediction model based on the comprehensive feature data set to predict the examination room resource data; An examination room resource allocation database construction module is used to construct an examination room resource allocation database according to the examination room resource data; The intelligent allocation module is used to intelligently allocate the examination time and examination location of testers in the next n time periods according to the examination room resource allocation database.

Images