A structured interview whole-process management and control method based on double-blind random matching

Abstract

This invention discloses a structured interview process control method based on double-blind random matching, belonging to the field of process control technology. This invention constructs a multi-level, three-dimensional constraint system for interviewer allocation, improving the accuracy and dimensionality of the constraints on matching examiners and candidates. It realizes an intelligent allocation mechanism considering multi-dimensional constraints, enhancing the fairness of allocation and the integrity of double-blind isolation. It can identify potential biases at each examiner node during the scoring process and systematic scoring deviations across examination rooms. It can accurately correct and dynamically compensate for the scoring deviation of each abnormal scoring examiner node by converting it into a scoring weight decay coefficient through a scoring weight decay mapping function, achieving real-time correction of scoring results. Under the premise of ensuring the unpredictability and tamper-proofness of the allocation process, it ensures the credibility of double-blind random matching, guaranteeing the compliance and sustainability of the system.

Description

Technical Field

[0001] This invention relates to the field of process control technology, specifically to a method for full-process control of structured interviews based on double-blind random matching. Background Technology

[0002] With the continuous deepening of talent selection mechanism reform and the continuous expansion of recruitment scale, structured interviews, as the core link of talent assessment and recruitment, directly affect the credibility of recruitment results and the accuracy of talent selection due to their fairness, standardization, and scientific nature. The entire interview process involves multiple key control nodes such as identity verification, personnel allocation, score aggregation, and anomaly detection. Deficiencies in the control of any link may lead to deviations or disputes in the interview results. Therefore, it is necessary to construct a structured interview full-process control method based on a double-blind random matching mechanism to achieve accurate personnel allocation and intelligent score anomaly detection under multi-dimensional constraints.

[0003] In existing technologies, most interview management systems rely on single-dimensional identity verification methods and manually arranged test room allocation modes. These methods cannot cover the cross-modal fusion identity verification needs of interviewees under multiple biometric modalities. They are also difficult to achieve in-depth identification and quantitative analysis of potential background relationships between examiners and candidates. In large-scale recruitment scenarios involving multiple test rooms and multiple positions, simply using random selection or manual arrangement for test room allocation can easily lead to problems such as omission of avoidance constraints, uneven load distribution, and failure of double-blind isolation when faced with a large number of examiner and candidate combinations across test rooms.

[0004] Meanwhile, existing technologies lack the ability to detect and dynamically correct multi-level anomalies in examiner scoring data, and lack an active identification mechanism for cross-examination room scoring consistency deviations. They cannot effectively avoid the distortion of scoring results caused by individual examiners' subjective biases or scoring drift. Furthermore, existing systems often rely on post-examination manual review for scoring correction, which has shortcomings such as delayed feedback, poor timeliness of correction, and limited intervention methods. In terms of perceiving scoring anomalies and quantifying scoring deviations, they cannot capture the examiners' true scoring tendencies and potential scoring deviations through in-examination room scoring dispersion analysis and cross-examination room scoring mean deviation analysis. This leads to a significant deviation between the final score data and the candidates' actual performance. At the same time, they lack the ability to constrain analysis based on background correlation conflict matrices, and cannot identify implicit interest relationships and potential fairness risks between examiners and candidates. They also cannot achieve unpredictable and tamper-proof double-blind random matching and allocation based on the aggregated random number seed sequence generated by the distributed consensus mechanism. Summary of the Invention

[0005] The purpose of this invention is to provide a structured interview process control method based on double-blind random matching, which solves the problems existing in the background technology.

[0006] To solve the above technical problems, the present invention adopts the following technical solution: The present invention provides a structured interview process control method based on double-blind random matching, including: Step 1: Obtain biometric data of each interviewee, perform cross-modal identity verification on each interviewee, thereby determining each examiner node and each candidate node, and establishing a background association conflict matrix between each examiner node and each candidate node accordingly.

[0007] Step 2: Based on the background association conflict matrix between each examiner node and each candidate node, construct a constraint relationship diagram for the allocation of personnel and examination room resources.

[0008] Step 3: Obtain the aggregated random number seed sequence, and perform a layer-by-layer double-blind random permutation and allocation operation based on the aggregated random number seed sequence and the personnel and examination room resource allocation constraint diagram. Then, perform load balancing correction and double-blind isolation integrity verification to generate the interview reporting personnel allocation plan.

[0009] Step 4: Based on the personnel allocation plan for interview registration, construct the set of extreme value removal weighted aggregation parameters for each examination room resource node. According to the personnel allocation plan for interview registration, transmit the allocation information and the set of extreme value removal weighted aggregation parameters corresponding to each examination room resource node to the scoring terminal device of each examination room through an encrypted channel. Collect the talent assessment scoring input data of each examiner node in real time, and apply the set of extreme value removal weighted aggregation parameters to the scoring input data to generate the real-time comprehensive scoring results of each candidate node.

[0010] Step 5: Based on the real-time comprehensive scoring results of each candidate node, perform cross-examination room scoring consistency anomaly detection and dynamic correction to generate the final score data for talent assessment and recruitment interviews.

[0011] The beneficial effects of this invention are as follows: This invention constructs a multi-level, three-dimensional constraint system for interviewer allocation, improving the accuracy and dimensionality of the matching constraints between examiners and candidates, forming a more rigorous and accurate constraint relationship diagram for the allocation of personnel and examination room resources. This provides a solid constraint foundation for subsequent layer-by-layer double-blind random permutation allocation operations, realizing an intelligent allocation mechanism that considers multiple constraints. Simultaneously, it considers multiple constraint factors such as the upper limit of candidate capacity, the upper limit of examiner configuration, personnel avoidance constraint distance, and conflict isolation constraint pairs, improving the fairness of allocation and the integrity of double-blind isolation. By constructing a cross-examination room scoring consistency anomaly detection mechanism, and introducing a two-level detection strategy of identifying abnormal scoring examiner nodes within the examination room and abnormal examination room resource nodes at the examination room level, it can identify each examiner... The system addresses potential biases in the scoring process and systematic scoring deviations across examination rooms. It not only analyzes scoring dispersion within a single examination room but also constructs a comprehensive scoring quality monitoring system that includes analysis of mean deviations across examination rooms. Based on a dynamic adjustment technique using extremum-removing weighted aggregation parameters, it accurately corrects and dynamically compensates for the scoring deviations of each abnormal scoring examiner node by converting them into scoring weight decay coefficients through a scoring weight decay mapping function. This enables real-time correction of scoring results. Furthermore, the system ensures the credibility of double-blind random matching by using a distributed consensus mechanism to generate aggregated random number seed sequences and a commitment hash value broadcast verification mechanism, guaranteeing the unpredictability and immutability of the allocation process. This ensures the system's compliance and sustainability. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Reference Figure 1As shown, the present invention provides a structured interview process control method based on double-blind random matching, including: Step 1: Obtain biometric data of each interviewee, perform cross-modal identity verification on each interviewee, thereby determining each examiner node and each candidate node, and establishing a background association conflict matrix between each examiner node and each candidate node accordingly.

[0016] In one specific embodiment, the biometric data of each interviewee is obtained by means of: obtaining the biometric data of each interviewee through facial detection and examination room check-in system.

[0017] It should be noted that the biometric data collected includes: feature sub-vectors of each biometric modality.

[0018] In a specific embodiment of the present invention, cross-modal identity verification is performed on each interviewee to determine each examiner node and each candidate node. The specific method is as follows: based on the biometric data of each interviewee, feature sub-vectors of each biometric modality of each interviewee are extracted, and each feature sub-vector is normalized and mapped to obtain the multi-dimensional fusion feature vector of each interviewee.

[0019] In one specific embodiment, a normalization mapping is performed to obtain the multidimensional fusion feature vector of each interviewee. The specific method is as follows: based on the feature sub-vectors of each biometric modality of each interviewee, the mean of each feature sub-vector is subtracted from each component value and then divided by the standard deviation of the feature sub-vector to ensure that the feature sub-vectors of each biometric modality are on a uniform numerical scale. The preset modality weights corresponding to each biometric modality are obtained from the local database. The normalized feature sub-vectors are multiplied by the corresponding preset modality weights and then concatenated sequentially to obtain the multidimensional fusion feature vector of each interviewee.

[0020] Retrieve the pre-stored baseline features and role attributes of each interviewee from the local database, perform cross-modal fusion comparison between the multi-dimensional fusion feature vector of each interviewee and the corresponding baseline features, and calculate the identity confidence score of each interviewee.

[0021] In one specific embodiment, the confidence score of each interviewee is calculated as follows: based on the multidimensional fusion feature vector and the corresponding baseline feature of each interviewee, the multidimensional fusion feature vector and the baseline feature are multiplied element-wise and summed to obtain the inner product value. The magnitude of the multidimensional fusion feature vector and the magnitude of the baseline feature are calculated separately. The inner product value is divided by the product of the two magnitudes to obtain the cosine similarity value. This cosine similarity value is the confidence score of each interviewee. The closer the value is to 1, the higher the degree of matching between the interviewee's biometrics and the pre-stored baseline features.

[0022] It should be noted that the roles include: examiner and examinee.

[0023] The system retrieves a preset verification threshold from the local database and marks interviewees whose identity confidence scores exceed the preset verification threshold as having passed verification. Based on the role attributes of each interviewee who has passed verification, the system divides them into examiner nodes and candidate nodes.

[0024] In a specific embodiment of the present invention, a background association conflict matrix is ​​established between each examiner node and each candidate node. The specific method is as follows: the graduating institution identifier, employing unit identifier, and cooperation relationship identifier of each examiner node and the graduating institution identifier and internship unit identifier of each candidate node are obtained from the local database, and an overlap analysis is performed accordingly to obtain the background association conflict intensity value between each examiner node and each candidate node. These values ​​are then arranged in a matrix form to establish the background association conflict matrix between each examiner node and each candidate node.

[0025] In one specific embodiment, the background association conflict strength value between each examiner node and each candidate node is obtained by the following method: based on the graduation school identifier, employing unit identifier, and cooperation relationship identifier of each examiner node obtained from the local database, and the graduation school identifier and internship unit identifier of each candidate node, the examiner node and each candidate node are compared one by one to see if they are the same in graduation school identifier, the same in employing unit identifier and internship unit identifier, and whether there is a relationship in cooperation relationship identifier. If they are the same or have a relationship, the overlap degree of that dimension is recorded as one; otherwise, it is recorded as zero. The overlap degrees of the three dimensions are added together to obtain the background association conflict strength value between each examiner node and each candidate node.

[0026] Step 2: Based on the background association conflict matrix between each examiner node and each candidate node, construct a constraint relationship diagram for the allocation of personnel and examination room resources.

[0027] In a specific embodiment of the present invention, a constraint relationship diagram for the allocation of personnel and examination room resources is constructed. The specific method is as follows: obtain the upper limit of the number of candidates and the upper limit of the number of examiners for each examination room resource node from the local database.

[0028] The personnel avoidance constraint distance is obtained from the local database based on the background association conflict intensity value of each examiner node and each candidate node. The personnel avoidance constraint distance between each examiner node and each candidate node is then calculated based on the background association conflict intensity value between each examiner node and each candidate node.

[0029] In one specific embodiment, the personnel avoidance constraint distance between each examiner node and each candidate node is calculated by multiplying the background association conflict intensity value between each examiner node and each candidate node with the personnel avoidance constraint distance corresponding to the unit background association conflict intensity value, and then calculating the personnel avoidance constraint distance between each examiner node and each candidate node.

[0030] Based on the background association conflict strength value between each examiner node and each candidate node, the conflict strength constraint threshold is obtained from the local database. Each examiner and candidate node pair whose background association conflict strength value exceeds the conflict strength constraint threshold is marked as a conflict isolation constraint pair, thus obtaining each conflict isolation constraint pair.

[0031] Based on the maximum capacity of candidates and the maximum number of examiners at each examination room resource node, the personnel avoidance constraint distance between each examiner node and each candidate node, and each conflict isolation constraint pair, a constraint relationship diagram between personnel and examination room resources is generated.

[0032] Step 3: Obtain the aggregated random number seed sequence, and perform a layer-by-layer double-blind random permutation and allocation operation based on the aggregated random number seed sequence and the personnel and examination room resource allocation constraint diagram. Then, perform load balancing correction and double-blind isolation integrity verification to generate the interview reporting personnel allocation plan.

[0033] In one specific embodiment, the method for obtaining the aggregated random number seed sequence is as follows: Multiple independent computing nodes in a distributed network generate their own randomly selected random number components based on their local timestamp and hardware entropy source. Each node calculates a commitment hash value for each randomly selected random number component and broadcasts these commitment hash values ​​to all other computing nodes in the distributed network. After all computing nodes have completed broadcasting their commitment hash values, each computing node publishes the plaintext of its randomly selected random number component. Each other computing node recalculates the hash value of the received plaintext of the randomly selected random number component and compares it with the previously received commitment hash value. If they match, the random number component of that computing node passes consensus verification. An XOR aggregation operation is then performed on all valid random number components that have passed consensus verification, thereby generating an unpredictable and tamper-proof aggregated random number seed sequence.

[0034] In a specific embodiment of the present invention, a layer-by-layer double-blind random permutation and allocation operation is performed. The specific method is as follows: based on the aggregated random number seed sequence, each candidate node is randomly arranged and permuted, and each candidate node is allocated to each examination room resource node in the order of permutation.

[0035] In one specific embodiment, each candidate node is assigned to each examination room resource node in sequence according to the replacement order. The specific method is as follows: the candidate node with the first replacement order is assigned to the examination room resource node with the sequence number 1, the candidate node with the second replacement order is assigned to the examination room resource node with the sequence number 2, and so on, until all examination room resource nodes have at least one person. Then, the assignment starts again from the examination room resource node with the sequence number 1, until the assignment is completed.

[0036] Based on different segments of the aggregated random number seed sequence, each examiner node is independently and randomly permuted. Under the premise of meeting the upper limit of examiner configuration and the personnel avoidance constraint distance between the examiner node and each candidate node already assigned to the same examination room resource node meets the requirements and is not subject to conflict isolation constraint, each chief examiner node is assigned to each examination room resource node. Similarly, random permutation and allocation are performed on each deputy examiner node in the set of deputy examiner nodes, thereby generating the initial assignment mapping result for the interview.

[0037] In a specific embodiment of the present invention, load balancing correction and double-blind isolation integrity verification are performed to generate an interview registration personnel allocation scheme. The specific method is as follows: based on the initial interview allocation mapping results, the balance deviation value of each examination room resource node is calculated, a preset balance threshold is obtained from the local database, each overloaded examination room resource node and each underloaded examination room resource node are screened, and an available examiner node or candidate node is selected from each overloaded examination room resource node and reassigned to each underloaded examination room resource node that meets the constraints, until the deviation value of all examination room resource nodes does not exceed the preset balance threshold.

[0038] In one specific embodiment, the equilibrium deviation value of each examination room resource node is calculated by the following method: based on the initial allocation mapping result of the interview, the deviation value between the actual number of candidates and the upper limit of the candidate capacity and the deviation value between the actual number of examiners and the upper limit of the examiner configuration of each examination room resource node are calculated, and these are added together to obtain the equilibrium deviation value of each examination room resource node.

[0039] In one specific embodiment, the method for filtering each overloaded examination room resource node and each underloaded examination room resource node is as follows: if the balance deviation value of a certain examination room resource node is greater than the preset balance threshold, then the examination room resource node is marked as an overloaded examination room resource node, otherwise it is marked as an underloaded examination room resource node, thereby filtering out each overloaded examination room resource node and each underloaded examination room resource node.

[0040] For the load-balanced corrected allocation results, iterate through all the pairing relationships between examiner nodes and candidate nodes in each examination room resource node, verify that the personnel avoidance constraint distance between examiner nodes and candidate nodes in each pairing relationship meets the requirements and that the pairing relationship does not belong to the conflict isolation constraint pair. If all pairing relationships pass the verification, the current allocation result is used as the allocation scheme for interview registration personnel. If there are pairing relationships that fail the verification, the examiner nodes or candidate nodes that fail the verification are re-randomly permuted and allocated based on the aggregated random number seed sequence until all pairing relationships pass the verification, thereby generating the allocation scheme for interview registration personnel.

[0041] Step 4: Based on the personnel allocation plan for interview registration, construct the set of extreme value removal weighted aggregation parameters for each examination room resource node. According to the personnel allocation plan for interview registration, transmit the allocation information and the set of extreme value removal weighted aggregation parameters corresponding to each examination room resource node to the scoring terminal device of each examination room through an encrypted channel. Collect the talent assessment scoring input data of each examiner node in real time, and apply the set of extreme value removal weighted aggregation parameters to the scoring input data to generate the real-time comprehensive scoring results of each candidate node.

[0042] In a specific embodiment of the present invention, a set of extreme value removal weighted aggregation parameters for each examination room resource node is constructed. The specific method is as follows: obtain the extreme value removal ratio parameter and the basic score weight coefficient corresponding to each examiner node from the local database.

[0043] It should be noted that the extreme value removal ratio parameter refers to the proportion of score data that falls within the highest and lowest score ranges before aggregating the score input data of each examiner node. This is used to eliminate the impact of abnormally extreme scores given by individual examiners due to subjective bias on the candidate's overall score. The basic score weight coefficient corresponding to each examiner node is a pre-set score influence parameter based on the role attributes of each examiner node, reflecting the difference in importance of examiners with different roles in score aggregation. For example, the basic score weight coefficient of the chief examiner node is higher than that of the deputy examiner node, so that the chief examiner's score accounts for a larger proportion in the overall score.

[0044] Based on the personnel allocation plan for the interview, the examiner combination structure of each examination room resource node is extracted, the number of examiner nodes in each examination room resource node is determined, and based on the score extreme value removal ratio parameter and the number of examiner nodes in each examination room resource node, the highest and lowest number of scores to be removed in each examination room resource node during the aggregation operation are calculated.

[0045] In one specific embodiment, the method for calculating the highest and lowest number of scores to be removed from each examination room resource node during aggregation is as follows: Based on the number of examiner nodes extracted from each examination room resource node according to the interviewee allocation plan, the number of examiner nodes is multiplied by the score extreme value removal ratio parameter and then rounded down to obtain the highest and lowest number of scores to be removed from each examination room resource node during aggregation. For example, if there are seven examiner nodes in a certain examination room resource node and the score extreme value removal ratio parameter is 15%, then seven multiplied by 15% equals 1.05, which is rounded down to one. That is, the highest number of scores to be removed from this examination room resource node is one, and the lowest number of scores is one.

[0046] Based on the basic scoring weight coefficients of each examiner node within each examination room resource node, we analyze the normalized scoring weight rules of examiner nodes within each examination room resource node after removing extreme scores.

[0047] In one specific embodiment, the normalized scoring weight rules for examiner nodes within each examination room resource node after removing extreme scores are analyzed. The specific method is as follows: Based on the basic scoring weight coefficients corresponding to each examiner node within each examination room resource node, after removing the examiner nodes corresponding to the highest and lowest scores, the basic scoring weight coefficients of the remaining examiner nodes are extracted. The sum of the basic scoring weight coefficients of the remaining examiner nodes is calculated. The basic scoring weight coefficients of each remaining examiner node are divided by the sum of these basic scoring weight coefficients to obtain the normalized scoring weights of each examiner node within each examination room resource node after removing extreme scores, ensuring that the sum of the normalized scoring weights of all remaining examiner nodes within each examination room resource node is equal to one. Based on the highest score removal quantity, lowest score removal quantity, and normalized score weighting rules of each examination room resource node, a set of extreme value removal weighted aggregation parameters is constructed for each examination room resource node.

[0048] In a specific embodiment of the present invention, the real-time comprehensive scoring results of each candidate node are generated by the following method: based on the scoring input data of each examiner node in each examination room resource node, the scoring input data of each candidate node in each examination room resource node at each examiner node is mapped, and the data are sorted according to their numerical values. Based on the extreme value removal weighted aggregation parameter set of each examination room resource node, the scoring data that are in the highest and lowest scoring ranges after sorting are removed. The remaining scoring input data of each examiner node are weighted and aggregated according to the normalized scoring weight rules to generate the real-time comprehensive scoring results of each candidate node.

[0049] In one specific embodiment, a weighted aggregation operation is performed to generate the real-time comprehensive score result for each candidate node. The specific method is as follows: After sorting the score input data of each candidate node in each examination room resource node according to their numerical values, the score data in the highest and lowest score ranges are removed. The remaining score input data of each examiner node and their corresponding normalized score weights are extracted. The score input data of each remaining examiner node is multiplied by its corresponding normalized score weight, and all products are summed to obtain the real-time comprehensive score result for that candidate node in that examination room resource node. This process is repeated for all candidate nodes within each examination room resource node to generate the real-time comprehensive score result for all candidate nodes.

[0050] Step 5: Based on the real-time comprehensive scoring results of each candidate node, perform cross-examination room scoring consistency anomaly detection and dynamic correction to generate the final score data for talent assessment and recruitment interviews.

[0051] In a specific embodiment of the present invention, cross-examination room scoring consistency anomaly detection and dynamic correction are performed to generate the final score data of talent assessment and recruitment interviews. The specific method is as follows: based on the scoring input data of each examiner node to each candidate node in each examination room resource node, the average score of each candidate node in each examination room resource node is calculated, and the scoring deviation value between the scoring input data of each examiner node to each candidate node in each examination room resource node and the average score is calculated accordingly. The scoring deviation threshold and the abnormal scoring ratio threshold are obtained from the local database. The ratio of the number of times the scoring deviation value of each examiner node exceeds the scoring deviation threshold to its total number of scoring times is counted. Each examiner node whose ratio exceeds the abnormal scoring ratio threshold is marked as an abnormal scoring examiner node.

[0052] Calculate the average score of each examination room resource node, calculate the cross-examination room score mean deviation between the average score of each examination room resource node and the total average score of all examination room resource nodes, obtain the cross-examination room deviation threshold from the local database, and mark each examination room resource node whose cross-examination room score mean deviation exceeds the cross-examination room deviation threshold as an abnormal examination room resource node.

[0053] Based on the scoring deviation values ​​of each abnormal scoring examiner node, the normalized scoring weights corresponding to each abnormal scoring examiner node in each examination room resource node are dynamically adjusted. Based on the cross-examination room scoring mean deviation of each abnormal examination room resource node, the scoring calibration compensation value of the corresponding examination room resource node is calculated. Based on the adjusted normalized scoring weights and scoring calibration compensation values, the real-time comprehensive scoring results of each candidate node are re-aggregated and calculated to generate the final score data of talent assessment and recruitment interview.

[0054] In one specific embodiment, the real-time comprehensive scoring results of each candidate node are re-aggregated and calculated to generate the final score data of the talent assessment and recruitment interview. The specific method is as follows: based on the method for generating the real-time comprehensive scoring results of each candidate node, the final score data of each candidate node is generated in the same way and combined into the final score data of the talent assessment and recruitment interview.

[0055] In a specific embodiment of the present invention, the normalized scoring weights corresponding to each abnormal scoring examiner node of each examination room resource node are dynamically adjusted. The specific method is as follows: the ratio of the number of times the scoring deviation value of each examiner node exceeds the scoring deviation threshold to its total number of scoring times is taken as the scoring deviation degree of each abnormal scoring examiner node.

[0056] The scoring weight decay mapping function is obtained from the local database. The degree of scoring deviation of each abnormal scoring examiner node is substituted into the scoring weight decay mapping function to obtain the scoring weight decay coefficient corresponding to each abnormal scoring examiner node.

[0057] It should be noted that the rating weight decay mapping function is a function that maps the degree of rating deviation of each abnormal rating examiner node to the corresponding rating weight decay coefficient. The rating weight decay coefficient is a value between zero and one. The characteristic of the rating weight decay mapping function is that the greater the degree of rating deviation, the smaller the output rating weight decay coefficient. That is, the more serious the rating deviation of the examiner, the more the rating weight is compressed. When the degree of rating deviation is close to zero, the decay coefficient is close to one, indicating that the weight is basically unchanged. When the degree of rating deviation is close to one, the decay coefficient approaches zero, indicating that the rating weight of the examiner has been greatly compressed.

[0058] The normalized score weights of each abnormal scorer node in the set of extreme value weighting aggregation parameters are multiplied by the corresponding score weight decay coefficient to obtain the adjusted normalized score weights.

[0059] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0060] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A double-blind random matching-based structured interview whole-process management and control method, characterized in that, include: Step 1: Obtain biometric data of each interviewee, perform cross-modal identity verification for each interviewee, thereby determining each examiner node and each candidate node, and establish a background association conflict matrix between each examiner node and each candidate node accordingly. Step 2: Based on the background association conflict matrix between each examiner node and each candidate node, construct a constraint relationship diagram for the allocation of personnel and examination room resources; Step 3: Obtain the aggregated random number seed sequence, perform a layer-by-layer double-blind random permutation and allocation operation based on the aggregated random number seed sequence and the personnel and examination room resource allocation constraint diagram, and perform load balancing correction and double-blind isolation integrity verification to generate the interview reporting personnel allocation plan; Step 4: Based on the allocation plan for interviewees, construct a set of extreme value removal weighted aggregation parameters for each examination room resource node. Based on the allocation plan for interviewees, transmit the allocation information and the set of extreme value removal weighted aggregation parameters corresponding to each examination room resource node to the scoring terminal device of each examination room through an encrypted channel. Collect the talent assessment scoring input data of each examiner node in real time, and apply the set of extreme value removal weighted aggregation parameters to the scoring input data to generate the real-time comprehensive scoring results of each candidate node. Step 5: Based on the real-time comprehensive scoring results of each candidate node, perform cross-examination room scoring consistency anomaly detection and dynamic correction to generate the final score data for talent assessment and recruitment interviews.

2. The structured interview full-process management and control method based on double-blind random matching according to claim 1, characterized in that, The specific method for performing cross-modal identity verification on each interviewee to determine the examiner and candidate nodes is as follows: Based on the biometric data collected from each interviewee, feature vectors of each biometric modality of each interviewee are extracted, and each feature vector is normalized and mapped to obtain a multidimensional fusion feature vector of each interviewee. Retrieve the pre-stored baseline features and role attributes of each interviewee from the local database, perform cross-modal fusion comparison between the multi-dimensional fusion feature vector of each interviewee and the corresponding baseline features, and calculate the identity confidence score of each interviewee. The system retrieves a preset verification threshold from the local database and marks interviewees whose identity confidence scores exceed the preset verification threshold as having passed verification. Based on the role attributes of each interviewee who has passed verification, the system divides them into examiner nodes and candidate nodes.

3. The method according to claim 1, wherein, The specific method for establishing the background association conflict matrix between each examiner node and each candidate node is as follows: The system retrieves the graduation institution identifier, employing unit identifier, and cooperation relationship identifier of each examiner node, and the graduation institution identifier and internship unit identifier of each candidate node from the local database. Based on this, it performs an overlap analysis to obtain the background association conflict intensity value between each examiner node and each candidate node. The values ​​are then arranged in matrix form to establish the background association conflict matrix between each examiner node and each candidate node.

4. The structured interview full-process management and control method based on double-blind random matching according to claim 3, characterized in that, The specific method for constructing the constraint relationship diagram between personnel and examination room resource allocation is as follows: Retrieve the maximum number of candidates and the maximum number of examiners for each examination room resource node from the local database; The personnel avoidance constraint distance corresponding to the background association conflict intensity value of each unit is obtained from the local database. Based on the background association conflict intensity value between each examiner node and each candidate node, the personnel avoidance constraint distance between each examiner node and each candidate node is calculated. Based on the background association conflict intensity value between each examiner node and each candidate node, the conflict intensity constraint threshold is obtained from the local database. Each examiner and candidate node pair whose background association conflict intensity value exceeds the conflict intensity constraint threshold is marked as a conflict isolation constraint pair, thereby obtaining each conflict isolation constraint pair. Based on the maximum capacity of candidates and the maximum number of examiners at each examination room resource node, the personnel avoidance constraint distance between each examiner node and each candidate node, and each conflict isolation constraint pair, a constraint relationship diagram between personnel and examination room resources is generated.

5. The method according to claim 4, wherein, The specific method for performing the layer-by-layer double-blind random permutation allocation operation is as follows: Based on the aggregated random number seed sequence, each candidate node is randomly arranged and replaced, and then each candidate node is assigned to each examination room resource node in the order of replacement. Based on different segments of the aggregated random number seed sequence, each examiner node is independently and randomly permuted. Under the premise of meeting the upper limit of examiner configuration and the personnel avoidance constraint distance between the examiner node and each candidate node already assigned to the same examination room resource node meets the requirements and is not subject to conflict isolation constraint, each chief examiner node is assigned to each examination room resource node. Similarly, random permutation and allocation are performed on each deputy examiner node in the set of deputy examiner nodes, thereby generating the initial assignment mapping result for the interview.

6. The structured interview full-process management and control method based on double-blind random matching according to claim 5, characterized in that, The specific method for performing load balancing correction and double-blind isolation integrity verification to generate an assignment scheme for interviewees is as follows: Based on the initial assignment mapping results of the interview, calculate the equilibrium deviation value of each examination room resource node, obtain the preset equilibrium threshold from the local database, filter each overloaded examination room resource node and each underloaded examination room resource node, select the reassignable examiner node or candidate node from each overloaded examination room resource node, and reassign them to each underloaded examination room resource node that meets the constraints, until the deviation value of all examination room resource nodes does not exceed the preset equilibrium threshold. For the load-balanced corrected allocation results, iterate through all the pairing relationships between examiner nodes and candidate nodes in each examination room resource node, verify that the personnel avoidance constraint distance between examiner nodes and candidate nodes in each pairing relationship meets the requirements and that the pairing relationship does not belong to the conflict isolation constraint pair. If all pairing relationships pass the verification, the current allocation result is used as the allocation scheme for interview registration personnel. If there are pairing relationships that fail the verification, the examiner nodes or candidate nodes that fail the verification are re-randomly permuted and allocated based on the aggregated random number seed sequence until all pairing relationships pass the verification, thereby generating the allocation scheme for interview registration personnel.

7. The method according to claim 1, wherein, The specific method for constructing the extreme value-weighted aggregation parameter set for each examination room resource node is as follows: Retrieve the extreme value removal ratio parameter and the basic score weight coefficient corresponding to each examiner node from the local database; Based on the personnel allocation plan for the interview, the examiner combination structure of each examination room resource node is extracted, the number of examiner nodes in each examination room resource node is determined, and based on the score extreme value removal ratio parameter and the number of examiner nodes in each examination room resource node, the highest and lowest number of scores to be removed in each examination room resource node during the aggregation operation are calculated. Based on the basic scoring weight coefficients of each examiner node within each examination room resource node, we analyze the normalized scoring weight rules of examiner nodes within each examination room resource node after removing extreme scores. Based on the highest score removal quantity, lowest score removal quantity, and normalized score weighting rules of each examination room resource node, a set of extreme value removal weighted aggregation parameters is constructed for each examination room resource node.

8. The method for full-process control of structured interviews based on double-blind random matching according to claim 7, characterized in that, The specific method for generating the real-time comprehensive score results for each candidate node is as follows: Based on the scoring input data of each examiner node in each examination room resource node, the scoring input data of each candidate node in each examination room resource node at each examiner node is mapped and sorted according to the numerical value. Based on the extreme value removal weighted aggregation parameter set of each examination room resource node, the scoring data in the highest and lowest scoring range after sorting is removed. The remaining scoring input data of each examiner node is weighted and aggregated according to the normalized scoring weight rule to generate the real-time comprehensive scoring result of each candidate node.

9. The method for full-process control of structured interviews based on double-blind random matching according to claim 1, characterized in that, The specific method for performing cross-examination room scoring consistency anomaly detection and dynamic correction to generate the final score data for talent assessment and recruitment interviews is as follows: Based on the scoring input data of each examiner node to each candidate node within each examination room resource node, the average score of each candidate node within each examination room resource node is calculated. Based on this, the scoring deviation value between the scoring input data of each examiner node to each candidate node within each examination room resource node and the average score is calculated. The scoring deviation threshold and abnormal scoring ratio threshold are obtained from the local database. The ratio of the number of times the scoring deviation value of each examiner node exceeds the scoring deviation threshold to its total number of scores is counted. Each examiner node whose ratio exceeds the abnormal scoring ratio threshold is marked as an abnormal scoring examiner node. Calculate the average score of each examination room resource node, calculate the cross-examination room score mean deviation between the average score of each examination room resource node and the total average score of all examination room resource nodes, obtain the cross-examination room deviation threshold from the local database, and mark each examination room resource node whose cross-examination room score mean deviation exceeds the cross-examination room deviation threshold as an abnormal examination room resource node. Based on the scoring deviation values ​​of each abnormal scoring examiner node, the normalized scoring weights corresponding to each abnormal scoring examiner node in each examination room resource node are dynamically adjusted. Based on the cross-examination room scoring mean deviation of each abnormal examination room resource node, the scoring calibration compensation value of the corresponding examination room resource node is calculated. Based on the adjusted normalized scoring weights and scoring calibration compensation values, the real-time comprehensive scoring results of each candidate node are re-aggregated and calculated to generate the final score data of talent assessment and recruitment interview.

10. The method for full-process control of structured interviews based on double-blind random matching according to claim 9, characterized in that, The specific method for dynamically adjusting the normalized scoring weights corresponding to each abnormal scoring examiner node in each examination room resource node is as follows: The ratio of the number of times the scoring deviation value of each examiner node exceeds the scoring deviation threshold to the total number of scoring times is used as the degree of scoring deviation of each abnormal scoring examiner node. The scoring weight decay mapping function is obtained from the local database. The degree of scoring deviation of each abnormal scoring examiner node is substituted into the scoring weight decay mapping function to obtain the scoring weight decay coefficient corresponding to each abnormal scoring examiner node. The normalized score weights of each abnormal scorer node in the set of extreme value weighting aggregation parameters are multiplied by the corresponding score weight decay coefficient to obtain the adjusted normalized score weights.

Images