Intelligent laboratory detection task allocation method and system based on big data

Abstract

The invention provides a laboratory detection task intelligent allocation method and system based on big data, and relates to the technical field of laboratory management, and the method comprises the steps: collecting detection task data, carrying out the detachable analysis to construct a collaborative flow chart, calculating a collaborative efficiency index and a task priority, and extracting personnel collaboration features to generate an association weight matrix; and allocating task queues by adopting a double-layer adaptive task allocation algorithm, performing density peak clustering on the first layer to divide task clusters, and optimizing an allocation scheme in combination with personnel ability and cooperative affinity on the second layer. According to the invention, the distribution efficiency and cooperation quality of complex laboratory detection tasks are improved.

Description

Technical Field

[0001] This invention relates to the field of laboratory management technology, and in particular to a method and system for intelligent allocation of laboratory testing tasks based on big data. Background Technology

[0002] In modern laboratory environments, with the increasing number and complexity of testing tasks, the efficient and rational allocation of these tasks has become a key issue in laboratory management. Traditional laboratory task allocation typically relies on manual experience or simple rotation systems, which are ill-suited to large-scale, highly complex testing demands. In recent years, with the development of big data technology, using data-driven methods for intelligent task allocation has become an important direction for improving laboratory efficiency.

[0003] Currently, the main techniques used in laboratory testing task allocation include basic workload balancing algorithms, simple expertise matching systems, and rudimentary considerations of collaboration. While these methods have improved the rationality of task allocation to some extent, they still have significant shortcomings when dealing with complex collaborative testing tasks.

[0004] The main defects and shortcomings of existing technologies include: Existing task allocation methods lack in-depth analysis of the decomposability of detection tasks, and cannot effectively decompose complex tasks into collaborative sub-tasks and construct reasonable task collaboration flowcharts, resulting in uneven resource allocation and low task execution efficiency.

[0005] Traditional allocation systems fail to fully explore and utilize historical collaboration data among testing personnel, cannot quantify the collaborative effectiveness of the testing team and the degree of synergy among personnel, and ignore the important impact of teamwork on testing quality and efficiency.

[0006] Existing technologies mostly adopt a single-dimensional allocation strategy, lacking a multi-level, adaptive task allocation algorithm framework. They cannot simultaneously consider task feature clustering and personnel matching optimization, making it difficult to achieve the optimal task allocation scheme in a dynamic detection environment. Summary of the Invention

[0007] This invention provides a method and system for intelligent allocation of laboratory testing tasks based on big data, which can solve the problems in the prior art.

[0008] A first aspect of this invention provides a method for intelligent allocation of laboratory testing tasks based on big data, comprising: Collect detection task data; The detection task is decomposed into multiple collaborative detection subtasks by performing a decomposability analysis. A task collaboration flowchart is constructed based on the dependencies between the subtasks to determine the execution order and data interaction nodes. Based on the task collaboration flowchart, the collaborative efficiency index of the detection team is calculated. Calculate the task allocation priority score based on the detection task data and the collaborative efficiency index, and generate a task queue to be allocated. Extract the collaboration pattern features of the inspectors from historical collaborative inspection records and generate a correlation weight matrix between the inspectors. Calculate the collaboration affinity between the inspectors based on the correlation weight matrix. The task queue is allocated based on a two-layer adaptive task allocation algorithm. In the first layer, task feature vectors are constructed, and the cosine similarity method is used to calculate the similarity matrix between tasks. Density peak clustering is performed based on the similarity matrix, and cluster centers are identified by local density and distance between tasks, dividing the tasks to be allocated into multiple task clusters. In the second layer, the task carrying capacity index of the detection personnel is obtained, and the matching degree matrix between the task clusters and the detection personnel is calculated by combining the cooperative affinity. The allocation scheme of each task cluster is optimized by a genetic algorithm, and the final scheme is obtained through iteration. The final solution is then sent to the testing personnel's terminal.

[0009] The detection task is decomposed into multiple collaborative detection subtasks based on a decomposability analysis. A task collaboration flowchart is constructed based on the dependencies between the subtasks to determine the execution order and data interaction nodes. Based on this task collaboration flowchart, the collaborative efficiency index of the detection team is calculated, including: The detection task is recursively decomposed into multiple collaborative detection subtasks. Each collaborative detection subtask is marked with a task flow node, and a task flow matrix is ​​generated. Based on the task flow matrix, the execution timing constraints of the collaborative detection subtasks are analyzed, and directed connections are established between collaborative detection subtasks with pre-determined dependencies to generate a task collaboration flowchart. Using collaborative detection subtasks as vertices, data transfer relationships between collaborative detection subtasks as edges, and data interaction volume as edge weights, the vertex degree and betweenness centrality are calculated; the critical path is determined based on the directed connection relationships in the task collaboration flowchart, and the importance of the data interaction node is calculated based on the weights of the critical path, the vertex degree, and the betweenness centrality. The collaborative execution efficiency is obtained by extracting the ratio of collaborative execution time to independent execution time from historical detection records, and the degree of parallel execution of collaborative detection subtasks is calculated based on the topological structure of the task collaboration flowchart. The collaborative efficiency index of the detection team is obtained by multiplying the importance of the data interaction node, the collaborative execution efficiency, and the degree of parallel execution.

[0010] The collaboration pattern features of the inspectors are extracted from historical collaborative inspection records, and a correlation weight matrix between the inspectors is generated. The collaboration affinity between the inspectors is calculated based on the correlation weight matrix, including: The collaboration duration feature, collaboration frequency feature, and quality score feature are extracted from historical collaborative detection records and normalized by maximum and minimum values ​​to obtain normalized feature values. Multiply the normalized average collaboration time by the duration weight coefficient, the normalized collaboration density by the frequency weight coefficient, and the normalized average quality score by the quality weight coefficient to obtain the multidimensional feature weight matrix. Calculate the L2 norm of the multidimensional feature weight matrix for each pair of inspectors, and divide the L2 norm by the maximum value of the L2 norm among all pairs of inspectors to obtain the collaboration strength coefficient; extract the quality fluctuation coefficient of each pair of inspectors and normalize it, and subtract the normalized quality fluctuation coefficient from 1 to obtain the collaboration stability coefficient; count the historical collaboration times of each pair of inspectors, and divide the collaboration times by the maximum value of the collaboration times among all pairs of inspectors to obtain the collaboration experience coefficient. The degree of synergy between the testing personnel is obtained by multiplying the synergy strength coefficient by the first comprehensive weighting coefficient, the synergy stability coefficient by the second comprehensive weighting coefficient, and the synergy experience coefficient by the third comprehensive weighting coefficient, and then summing the three weighted coefficients.

[0011] In the first layer, task feature vectors are constructed, and the cosine similarity method is used to calculate the inter-task similarity matrix. Density peak clustering is performed based on the similarity matrix, and cluster centers are identified through local density and inter-task distance, dividing the tasks to be assigned into multiple task clusters, including: Extract the feature parameters of the detection task and multiply them by the corresponding feature weight coefficients to obtain the task feature vector; Calculate the inner product and norm product of the feature vectors of two detection tasks, divide the inner product by the norm product to obtain the similarity between the two detection tasks, and traverse all detection task pairs to generate a task similarity matrix. Subtract the similarity in the task similarity matrix from 1 to obtain the task distance matrix; select the maximum distance value in the task distance matrix, and multiply the maximum distance value by the density truncation coefficient to obtain the truncation distance; count the number of tasks whose distance from each detection task to other detection tasks is less than the truncation distance, and obtain the local density of the detection task. Calculate the distance between each detection task and the first detection task, and select the smallest distance value as the minimum distance of the first detection task; calculate the distance between the detection task with the highest local density and all other detection tasks, and select the largest distance value as the minimum distance of that detection task; repeat the above process for each detection task to obtain its minimum distance, and multiply the local density of each detection task by its corresponding minimum distance to obtain the decision value of that detection task. The local maximum value among the decision values ​​is selected as the cluster center, and the remaining detection tasks are assigned to the nearest cluster center to form multiple task clusters.

[0012] In the second layer, the task capacity index of the testing personnel is obtained, and the matching degree matrix between task clusters and testing personnel is calculated by combining the cooperative affinity. The allocation scheme of each task cluster is optimized using a genetic algorithm, and the final scheme is obtained through iteration, including: The task carrying capacity index is obtained based on the historical tasks of the testing personnel; Extract the success rate of inspectors completing tasks of different complexities from historical inspection records, and fit the success rate with the corresponding task complexity to obtain the complexity fitness. The average complexity of all tasks within a task cluster is calculated to obtain the cluster-wide task complexity. The task carrying capacity index is multiplied by a first balance coefficient, and the cooperative affinity is multiplied by a second balance coefficient. The first and second balance coefficients are added together, and the difference between the cluster-wide task complexity and the complexity fitness is subtracted and multiplied by a third balance coefficient to obtain the matching degree matrix between the testing personnel and the task cluster. The fitness value is obtained by calculating the sum of the matching degrees of all inspectors and their assigned task clusters based on the matching degree matrix and subtracting the constraint violation penalty term. Construct a multi-genomic chromosome, with each gene locus corresponding to a task cluster. The gene locus value is assigned to the detection personnel number of that task cluster. Sort the chromosomes according to their fitness values ​​from largest to smallest. Select chromosomes according to the sorting and perform partial mapping crossover operations, exchanging the values ​​of gene loci within the crossover region. Randomly select two gene loci in the chromosomes and exchange their corresponding detection personnel numbers to complete the mutation operation. Repeat the crossover operation and the mutation operation until the maximum number of iterations is reached. Select the chromosome code corresponding to the highest fitness value as the final task allocation scheme.

[0013] Chromosomes are sorted from highest to lowest fitness value. Chromosomes are selected based on this sorting and subjected to partial mapping crossover. Gene loci values ​​are exchanged within the crossover region. Finally, two gene loci on randomly selected chromosomes are swapped with their corresponding personnel IDs to complete the mutation operation. Chromosomes are sorted from largest to smallest fitness value. The ratio of the current chromosome's fitness value to the population's optimal fitness value is calculated based on the fitness value to obtain the population selection ratio. Parent chromosomes and offspring chromosomes are selected based on the sorting and the population selection ratio. Two chromosomes are randomly selected from the chosen parent chromosomes for partial mapping crossover. The basic crossover coefficient is summed with the negative exponential function of the current iteration number and multiplied by the chromosome length to obtain the crossover region length. The one-to-one correspondence between the gene positions of the two parent chromosomes is recorded in the crossover region to form a mapping table. The detection personnel number of the second parent chromosome in the crossover region is copied to the corresponding position of the offspring chromosome. The duplicate detection personnel number in the offspring chromosome is eliminated according to the mapping table, and the exchange of gene position values ​​is completed. The task contribution of each gene locus in the offspring chromosome is calculated by summing the matching degree between the task cluster corresponding to each gene locus and the assigned testing personnel. The task contribution of each gene locus is subtracted from the maximum task contribution of all gene loci and normalized to obtain the mutation selection probability. Two mutation positions are determined according to the mutation selection probability. The testing personnel numbers corresponding to the two mutation positions are swapped to complete the mutation operation.

[0014] A second aspect of the present invention provides a big data-based intelligent allocation system for laboratory testing tasks, comprising: The first unit is used to collect data for the detection task; The second unit is used to perform a decomposability analysis on the detection task, decompose the task into multiple collaborative detection sub-tasks, construct a task collaboration flowchart based on the dependencies between sub-tasks, determine the execution order and data interaction nodes, and calculate the collaborative efficiency index of the detection team based on the task collaboration flowchart. The third unit is used to calculate the task allocation priority score based on the detection task data and the collaborative efficiency index, and generate a task queue to be allocated. The fourth unit is used to extract the collaboration pattern features of the inspectors from historical collaborative inspection records and generate a correlation weight matrix between the inspectors, and calculate the collaboration affinity between the inspectors based on the correlation weight matrix. The fifth unit is used to allocate the task queue based on a two-layer adaptive task allocation algorithm. In the first layer, task feature vectors are constructed, and the cosine similarity method is used to calculate the similarity matrix between tasks. Density peak clustering is performed based on the similarity matrix, and cluster centers are identified by local density and distance between tasks, dividing the tasks to be allocated into multiple task clusters. In the second layer, the task carrying capacity index of the detection personnel is obtained, and the matching degree matrix between the task clusters and the detection personnel is calculated by combining the cooperative affinity. The allocation scheme of each task cluster is optimized by using a genetic algorithm, and the final scheme is obtained through iteration. The sixth unit is used to distribute the final solution to the testing personnel's terminal.

[0015] A third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0016] Fourth aspect of the present invention, A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0017] The beneficial effects of the present invention are as follows: This invention achieves the scientific decomposition and collaborative execution of complex tasks by performing a decomposability analysis on the detection task and constructing a collaborative flowchart, thereby improving the overall work efficiency of the team and reducing the communication costs and coordination difficulties in task execution.

[0018] This invention introduces a collaborative affinity calculation mechanism based on historical collaboration data. By extracting the collaboration pattern features of the testing personnel and generating a correlation weight matrix, it can accurately identify the optimal collaboration combination, significantly enhance the synergy among team members, and improve the accuracy and consistency of the testing results.

[0019] This invention employs a two-layer adaptive task allocation algorithm. In the first layer, tasks are scientifically grouped through density peak clustering. In the second layer, tasks are optimized by combining the ability indicators of the testing personnel and their collaborative affinity. This effectively balances the workload, improves resource utilization, and ensures the rationality and fairness of task allocation, adapting to the complex and ever-changing testing environment in the laboratory. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the intelligent allocation method for laboratory testing tasks based on big data, as described in an embodiment of the present invention. Figure 2 This is a clustering decision graph for density peaks. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0022] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0023] Figure 1 This is a flowchart illustrating the intelligent allocation method for laboratory testing tasks based on big data, as described in an embodiment of the present invention. Figure 1 As shown, the method includes: Collect detection task data; The detection task is decomposed into multiple collaborative detection subtasks by performing a decomposability analysis. A task collaboration flowchart is constructed based on the dependencies between the subtasks to determine the execution order and data interaction nodes. Based on the task collaboration flowchart, the collaborative efficiency index of the detection team is calculated. Calculate the task allocation priority score based on the detection task data and the collaborative efficiency index, and generate a task queue to be allocated. Extract the collaboration pattern features of the inspectors from historical collaborative inspection records and generate a correlation weight matrix between the inspectors. Calculate the collaboration affinity between the inspectors based on the correlation weight matrix. The task queue is allocated based on a two-layer adaptive task allocation algorithm. In the first layer, task feature vectors are constructed, and the cosine similarity method is used to calculate the similarity matrix between tasks. Density peak clustering is performed based on the similarity matrix, and cluster centers are identified by local density and distance between tasks, dividing the tasks to be allocated into multiple task clusters. In the second layer, the task carrying capacity index of the detection personnel is obtained, and the matching degree matrix between the task clusters and the detection personnel is calculated by combining the cooperative affinity. The allocation scheme of each task cluster is optimized by a genetic algorithm, and the final scheme is obtained through iteration. The final solution is then sent to the testing personnel's terminal.

[0024] In one optional implementation, a decomposability analysis is performed on the detection task, breaking it down into multiple collaborative detection sub-tasks. A task collaboration flowchart is constructed based on the dependencies between the sub-tasks to determine the execution order and data interaction nodes. Based on the task collaboration flowchart, the collaborative efficiency index of the detection team is calculated, including: The detection task is recursively decomposed into multiple collaborative detection subtasks. Each collaborative detection subtask is marked with a task flow node, and a task flow matrix is ​​generated. Based on the task flow matrix, the execution timing constraints of the collaborative detection subtasks are analyzed, and directed connections are established between collaborative detection subtasks with pre-determined dependencies to generate a task collaboration flowchart. Using collaborative detection subtasks as vertices, data transfer relationships between collaborative detection subtasks as edges, and data interaction volume as edge weights, the vertex degree and betweenness centrality are calculated; the critical path is determined based on the directed connection relationships in the task collaboration flowchart, and the importance of the data interaction node is calculated based on the weights of the critical path, the vertex degree, and the betweenness centrality. The collaborative execution efficiency is obtained by extracting the ratio of collaborative execution time to independent execution time from historical detection records, and the degree of parallel execution of collaborative detection subtasks is calculated based on the topological structure of the task collaboration flowchart. The collaborative efficiency index of the detection team is obtained by multiplying the importance of the data interaction node, the collaborative execution efficiency, and the degree of parallel execution.

[0025] First, the inspection task is decomposed hierarchically. For example, a quality inspection task for an industrial product can be broken down into three primary sub-tasks: appearance inspection, functional inspection, and performance testing. Appearance inspection can be further decomposed into two secondary sub-tasks: dimensional measurement and surface defect inspection. Functional inspection can be decomposed into two secondary sub-tasks: circuit continuity testing and button function testing. Performance testing can be decomposed into two secondary sub-tasks: durability testing and stability testing. This recursive decomposition continues until each sub-task is simple enough to be completed by a single inspector or a team.

[0026] For each collaborative inspection subtask, a task flow node is labeled, generating a task flow matrix. The task flow node includes the task start point, task completion point, and intermediate data interaction points. In the example, the start point for appearance inspection is labeled A1, and the completion point is labeled A2; the start point for functional inspection is labeled B1, and the completion point is labeled B2; and the start point for performance testing is labeled C1, and the completion point is labeled C2. The task flow matrix is ​​a two-dimensional table, with rows and columns representing task flow nodes. Matrix elements represent the amount of data interaction from a row node to a column node. For example, if the amount of data interaction from the appearance inspection completion point A2 to the functional inspection start point B1 is 5 data items, then the value at the corresponding position in the task flow matrix is ​​5.

[0027] The execution timing constraints of collaborative detection subtasks are analyzed based on a task flow matrix. By analyzing the positions of non-zero elements in the matrix, the dependencies between subtasks are determined. For example, if the data interaction between A2 and B1 is not zero, it indicates that functional detection depends on the result of appearance detection and can only begin after appearance detection is completed. For subtasks with prior dependencies, directed connections are established, ultimately generating a task collaboration flowchart. In the example, a directed connection is established from appearance detection to functional detection, indicating that functional detection must be performed after appearance detection is completed.

[0028] Using collaborative detection subtasks as vertices, data transfer relationships between subtasks as edges, and data interaction volume as edge weights, graph theory-related metrics are calculated. Vertex degree reflects the number of connections between a subtask and other subtasks, representing the breadth of collaboration for that subtask. For example, if appearance detection interacts with both functional detection and performance testing, its degree is 2. Betweenness centrality measures the transitive role of a subtask between other subtask pairs, calculated by counting the number of shortest paths passing through that vertex. In the example, if most information transfer between subtasks requires passing through functional detection, then functional detection has a higher betweenness centrality.

[0029] Identify the critical path in the task collaboration flowchart, which is the longest path from project start to finish. Delays on the critical path will directly affect the overall project completion time. When calculating the importance of data interaction nodes, the edge weights, vertex degrees, and betweenness centrality are weighted and summed. The weight coefficients can be adjusted according to the actual situation; for example, the edge weight coefficient is 0.5, the vertex degree weight coefficient is 0.3, and the betweenness centrality weight coefficient is 0.2. For the functional detection in the example, assuming its total associated edge weights are 15, vertex degree is 3, and betweenness centrality is 0.4, its importance is calculated as 15 × 0.5 + 3 × 0.3 + 0.4 × 0.2 = 8.08.

[0030] Extract collaborative execution efficiency from historical detection records. For each subtask, calculate the ratio of its collaborative execution time to its independent execution time. A ratio less than 1 indicates that collaborative execution saved time, while a ratio greater than 1 indicates that collaboration incurred additional overhead. For example, if appearance detection requires 5 hours to execute independently and 4 hours to execute collaboratively, its collaborative execution efficiency is 4 / 5 = 0.8. Average the collaborative execution efficiency of all subtasks to obtain the overall collaborative execution efficiency.

[0031] Calculate the degree of parallel execution for collaborative detection subtasks. Based on the topology of the task collaboration flowchart, analyze the proportion of subtasks that can be executed simultaneously. Specifically, this can be achieved by calculating the ratio of the maximum number of parallel tasks to the total number of tasks. In the example, if a maximum of 3 out of 6 subtasks can be executed simultaneously, then the degree of parallel execution is 3 / 6 = 0.5.

[0032] The collaboration effectiveness index of the testing team is obtained by multiplying the importance of data interaction nodes, collaborative execution efficiency, and parallel execution degree. For example, if the average importance of data interaction nodes is 7.5, the collaborative execution efficiency is 0.85, and the parallel execution degree is 0.6, then the collaboration effectiveness index is 7.5 × 0.85 × 0.6 = 3.83. The higher this index, the better the collaboration effectiveness of the testing team.

[0033] The collaboration efficiency index can be used to quantitatively evaluate the collaboration level of a testing team, identify collaboration bottlenecks, and provide a basis for optimizing testing tasks. For example, if a testing team has a low collaboration efficiency index, it can be analyzed whether this is due to insufficient importance of data interaction nodes, low collaborative execution efficiency, or insufficient parallel execution, allowing for targeted improvements. When the collaboration efficiency index reaches a preset threshold (e.g., 4.0) or higher, the testing team can be considered to have good collaboration capabilities.

[0034] Furthermore, the collaborative effectiveness index of different testing teams can be compared horizontally to identify best practices and promote them among teams. Regularly measuring changes in the collaborative effectiveness index can also assess the development trend of team collaboration capabilities, providing support for management decisions.

[0035] In one optional implementation, the collaboration pattern features of the inspectors are extracted from historical collaborative inspection records, and a correlation weight matrix between the inspectors is generated. The collaborative affinity between the inspectors is calculated based on the correlation weight matrix, including: The collaboration duration feature, collaboration frequency feature, and quality score feature are extracted from historical collaborative detection records and normalized by maximum and minimum values ​​to obtain normalized feature values. Multiply the normalized average collaboration time by the duration weight coefficient, the normalized collaboration density by the frequency weight coefficient, and the normalized average quality score by the quality weight coefficient to obtain the multidimensional feature weight matrix. Calculate the L2 norm of the multidimensional feature weight matrix for each pair of inspectors, and divide the L2 norm by the maximum value of the L2 norm among all pairs of inspectors to obtain the collaboration strength coefficient; extract the quality fluctuation coefficient of each pair of inspectors and normalize it, and subtract the normalized quality fluctuation coefficient from 1 to obtain the collaboration stability coefficient; count the historical collaboration times of each pair of inspectors, and divide the collaboration times by the maximum value of the collaboration times among all pairs of inspectors to obtain the collaboration experience coefficient. The degree of synergy between the testing personnel is obtained by multiplying the synergy strength coefficient by the first comprehensive weighting coefficient, the synergy stability coefficient by the second comprehensive weighting coefficient, and the synergy experience coefficient by the third comprehensive weighting coefficient, and then summing the three weighted coefficients.

[0036] Extracting collaboration pattern features from historical collaborative detection records includes extracting collaboration duration features, collaboration frequency features, and quality score features. For two inspectors E1 and E2, assuming they jointly completed n collaborative detection tasks, with each collaboration lasting for durations t1, t2, ..., t... n If the collaboration duration is n hours, then their average collaboration time is the arithmetic mean of these n collaboration durations. For example, if E1 and E2 completed 3 collaborative detection tasks with durations of 4 hours, 5 hours, and 6 hours respectively, then their average collaboration time is 5 hours. Collaboration frequency can be represented as collaboration density, which is the number of times two inspectors jointly complete a detection task within a specific time period (e.g., one month). For example, if E1 and E2 jointly completed 10 detection tasks in the past month, and that month had 22 working days, then their collaboration density is 0.45 times / day. Quality score characteristic refers to the quality score of each collaborative detection task. Assuming the quality scores of n collaborations are q1, q2, ..., q... n The average quality score is the arithmetic mean of these n scores. For example, if the quality scores for three collaborations are 85, 90, and 95, the average quality score is 90.

[0037] The extracted features are normalized using maximum and minimum values ​​to ensure comparability of features with different dimensions. For example, if the longest average collaboration time among all testing personnel pairs is 10 hours and the shortest is 2 hours, while the average collaboration time between E1 and E2 is 5 hours, then the normalized value is (5-2) / (10-2) = 0.375. Similarly, collaboration density and average quality score are also normalized using this method.

[0038] The multidimensional feature weight matrix is ​​obtained by multiplying by the corresponding weight coefficients. Assuming the duration weight coefficient is 0.3, the frequency weight coefficient is 0.3, and the quality weight coefficient is 0.4, then the multidimensional feature weight matrices of E1 and E2 contain three weight values: the normalized average collaboration duration of 0.375 multiplied by 0.3 equals 0.1125; the normalized collaboration density of 0.5 multiplied by 0.3 equals 0.15; and the normalized average quality score of 0.8 multiplied by 0.4 equals 0.32.

[0039] Calculate the L2 norm of the multidimensional feature weight matrix for each pair of inspectors, which is the square root of the sum of the squares of the weight values. For E1 and E2, the L2 norm is: If the maximum value of the L2 norm for all testing personnel is 0.5, then the synergy strength coefficient between E1 and E2 is 0.365 / 0.5 = 0.73.

[0040] The quality fluctuation coefficient reflects the stability of the testing quality and can be represented by the standard deviation of the quality scores. Assuming the standard deviation of the quality scores from the three collaborations between E1 and E2 is 5, and the standard deviation of the quality scores from all testing personnel is between 1 and 10, then the normalized quality fluctuation coefficient is (5-1) / (10-1) = 0.444. The collaboration stability coefficient is 1 minus the normalized quality fluctuation coefficient, i.e., 1 - 0.444 = 0.556.

[0041] The collaboration experience coefficient is obtained by counting the number of times each pair of inspectors has collaborated in history and dividing that number by the maximum number of collaborations among all pairs of inspectors. If E1 and E2 have collaborated a total of 15 times and the maximum number of collaborations between inspector pairs is 30, then the collaboration experience coefficient is 15 / 30 = 0.5.

[0042] The synergy strength coefficient, synergy stability coefficient, and synergy experience coefficient are multiplied by their respective comprehensive weighting coefficients, and then summed to obtain the synergy affinity between the testing personnel. Assuming the first comprehensive weighting coefficient is 0.5, the second comprehensive weighting coefficient is 0.3, and the third comprehensive weighting coefficient is 0.2, then the synergy affinity between E1 and E2 is 0.73×0.5 + 0.556×0.3 + 0.5×0.2 = 0.6318.

[0043] In practical applications, a collaborative affinity matrix containing all testing personnel can be constructed based on historical collaborative testing records. For example, if there are 4 testing personnel, a 4×4 collaborative affinity matrix will be formed, where the diagonal elements (representing the collaborative affinity with themselves) can be set to 1 or ignored, and the off-diagonal elements represent the collaborative affinity between two different testing personnel.

[0044] The synergy affinity calculated using this implementation method can be used to optimize subsequent personnel allocation strategies. For example, when forming a new testing team, priority can be given to combinations of testing personnel with high synergy affinity to improve testing efficiency and quality. For newly joined testing personnel, due to the lack of historical collaboration data, a certain number of experimental collaboration tasks can be assigned first to accumulate initial collaboration data before gradually integrating them into the collaborative testing system.

[0045] To ensure the timeliness and accuracy of the calculation results, the co-affinity matrix can be updated periodically (e.g., weekly or monthly) based on the latest historical co-detection records. Simultaneously, a decay factor can be set to gradually reduce the influence of older collaboration records on the current co-affinity, thereby more accurately reflecting the current collaboration status of the testing personnel.

[0046] In one optional implementation, in the first layer, a task feature vector is constructed, a cosine similarity method is used to calculate the inter-task similarity matrix, density peak clustering is performed based on the similarity matrix, and cluster centers are identified through local density and inter-task distance, dividing the tasks to be assigned into multiple task clusters, including: Extract the feature parameters of the detection task and multiply them by the corresponding feature weight coefficients to obtain the task feature vector; Calculate the inner product and norm product of the feature vectors of two detection tasks, divide the inner product by the norm product to obtain the similarity between the two detection tasks, and traverse all detection task pairs to generate a task similarity matrix. Subtract the similarity in the task similarity matrix from 1 to obtain the task distance matrix; select the maximum distance value in the task distance matrix, and multiply the maximum distance value by the density truncation coefficient to obtain the truncation distance; count the number of tasks whose distance from each detection task to other detection tasks is less than the truncation distance, and obtain the local density of the detection task. Calculate the distance between each detection task and the first detection task, and select the smallest distance value as the minimum distance of the first detection task; calculate the distance between the detection task with the highest local density and all other detection tasks, and select the largest distance value as the minimum distance of that detection task; repeat the above process for each detection task to obtain its minimum distance, and multiply the local density of each detection task by its corresponding minimum distance to obtain the decision value of that detection task. The local maximum value among the decision values ​​is selected as the cluster center, and the remaining detection tasks are assigned to the nearest cluster center to form multiple task clusters.

[0047] In practice, the construction of the task feature vector is based on the feature parameters of the detection task. For each detection task, its feature parameters need to be extracted first. These feature parameters may include the coordinates of the detection area, the type of the detection target, the detection priority, and the execution time window. For example, for a task to detect drone intrusion, its feature parameters include: the detection area coordinates are (120.5, 35.6), the detection target type is "drone", the priority is 5, and the time window is 60 seconds. These feature parameters need to be multiplied by pre-set feature weight coefficients to reflect the importance of different features in the task similarity judgment. Assuming the weights for the above features are 0.3, 0.4, 0.2, and 0.1 respectively, the feature vector of the detection task is formed by multiplying and combining the corresponding feature values.

[0048] When calculating the similarity between tasks, the cosine similarity method is used. For two detection tasks D1 and D2, assuming their feature vectors are TD1 and TD2 respectively, the inner product of these two vectors is first calculated: corresponding elements of TD1 and TD2 are multiplied and summed. By traversing all task pairs, a complete task similarity matrix is ​​generated.

[0049] To construct the inter-task distance matrix, you need to subtract each element value in the similarity matrix from 1. For example, if the similarity between tasks D1 and D2 is 0.92, then the distance between them is 1 - 0.92 = 0.08. Perform this operation on all task pairs to obtain the inter-task distance matrix.

[0050] In density peak clustering, the first step is to determine the cutoff distance. Find the maximum distance value in the inter-task distance matrix, for example, 0.85, and multiply it by a preset density cutoff coefficient (e.g., 0.2) to obtain a cutoff distance of 0.17. For each detection task, count the number of tasks whose distances to other tasks are less than the cutoff distance; this number is used as the local density for that task. For example, if task D3's distances to all five other tasks are less than the cutoff distance, its local density is 5.

[0051] Calculating the minimum distance for each task involves different scenarios. For the first detection task, calculate its distance to all other detection tasks and select the minimum value as its minimum distance. For the task with the highest local density, calculate its distance to all other tasks and select the maximum value as its minimum distance. For the remaining tasks, identify all other tasks with a local density greater than the task's own, calculate the distance between the task and these tasks, and select the minimum value as the task's minimum distance. For example, if task D4 has a local density of 3, and there are two tasks D5 and D6 with local densities of 5 and 4 respectively, and the distances between D4 and D5 and D6 are 0.25 and 0.18 respectively, then the minimum distance for task D4 is 0.18.

[0052] The decision value is obtained by multiplying the local density of each task by its minimum distance. For example, if the local density of task D7 is 4 and the minimum distance is 0.3, then its decision value is 4 × 0.3 = 1.2. The local maximum value among all decision values ​​is selected as the cluster center. Specifically, the decision value of each task is compared with the decision values ​​of its neighboring tasks (tasks that are close in distance in the feature space). If the decision value of a task is greater than the decision values ​​of all its neighboring tasks, then that task is the cluster center.

[0053] For example, in a scenario with 10 detection tasks, the above calculation identifies tasks D8 (decision value 1.5), D9 (decision value 1.3), and D10 (decision value 1.4) as cluster centers. For the remaining tasks, the distances between them and each cluster center are calculated, and each task is assigned to the cluster containing the nearest cluster center. For instance, if the distances between task D11 and cluster centers D8, D9, and D10 are 0.15, 0.28, and 0.32 respectively, then task D11 is assigned to the cluster centered on task D8.

[0054] In this method, all detection tasks are divided into multiple task clusters, each containing a cluster center and several tasks with high similarity to that center. This task grouping method facilitates efficient task allocation and execution in subsequent processing.

[0055] Figure 2 This is a density peak clustering decision map, illustrating the distribution characteristics of detection tasks in the local density-minimum distance two-dimensional space and the cluster center identification process based on this invention. The X-axis represents local density, reflecting the number of similar tasks around each detection task, while the Y-axis represents minimum distance, the shortest distance from a task to a higher-density task. The product of these two values ​​constitutes the decision value, used for automatic cluster center identification. The three cluster centers marked by red dots in the upper right region exhibit high local density and high minimum distance, with decision values ​​of 23.8, 19.5, and 15.84, respectively, demonstrating the core advantage of this invention's algorithm in automatically identifying local maximum points as cluster centers. The blue dots represent 20 ordinary task points, distributed in a medium density and distance range, with relatively low decision values; these will be assigned to the nearest cluster center to form task clusters. The green dots represent 8 boundary task points, located in a low-density region, typically representing anomalous or independent detection tasks. The visualization results verify the effectiveness of the density peak clustering algorithm of this invention. Compared with traditional methods such as K-means, it can automatically determine the number of clusters without the need for manual parameter preset. At the same time, it realizes the scientific identification of cluster centers through the calculation of decision values, laying a solid technical foundation for subsequent intelligent task allocation.

[0056] In one optional implementation, in the second layer, the task capacity index of the inspectors is obtained, and the matching degree matrix between task clusters and inspectors is calculated by combining the cooperative affinity. A genetic algorithm is then used to optimize the allocation scheme for each task cluster, and the final scheme is obtained through iteration, including: The task carrying capacity index is obtained based on the historical tasks of the testing personnel; Extract the success rate of inspectors completing tasks of different complexities from historical inspection records, and fit the success rate with the corresponding task complexity to obtain the complexity fitness. The average complexity of all tasks within a task cluster is calculated to obtain the cluster-wide task complexity. The task carrying capacity index is multiplied by a first balance coefficient, and the cooperative affinity is multiplied by a second balance coefficient. The first and second balance coefficients are added together, and the difference between the cluster-wide task complexity and the complexity fitness is subtracted and multiplied by a third balance coefficient to obtain the matching degree matrix between the testing personnel and the task cluster. The fitness value is obtained by calculating the sum of the matching degrees of all inspectors and their assigned task clusters based on the matching degree matrix and subtracting the constraint violation penalty term. Construct a multi-genomic chromosome, with each gene locus corresponding to a task cluster. The gene locus value is assigned to the detection personnel number of that task cluster. Sort the chromosomes according to their fitness values ​​from largest to smallest. Select chromosomes according to the sorting and perform partial mapping crossover operations, exchanging the values ​​of gene loci within the crossover region. Randomly select two gene loci in the chromosomes and exchange their corresponding detection personnel numbers to complete the mutation operation. Repeat the crossover operation and the mutation operation until the maximum number of iterations is reached. Select the chromosome code corresponding to the highest fitness value as the final task allocation scheme.

[0057] When allocating tasks at the second level, it is necessary to obtain the task capacity index of each testing personnel. This index can be obtained by analyzing the historical task completion data of each testing personnel. For example, for a particular testing personnel, all task data completed within the past 6 months can be collected, including factors such as task completion time, task complexity, and testing quality. This data can then be weighted and averaged to form the testing personnel's task capacity index value. Assuming a testing center has 5 testing personnel, the task capacity indices obtained by analyzing their historical data are as follows: E1 = 0.85, E2 = 0.92, E3 = 0.76, E4 = 0.88, and E5 = 0.79.

[0058] By analyzing historical testing records, the success rate of each tester on tasks of varying complexity can be obtained. For example, historical data analysis of tester E1 shows a success rate of 0.85% on a task with a complexity of 0.6, 0.78% on a task with a complexity of 0.7, and 0.72% on a task with a complexity of 0.8. Fitting these discrete complexity-success rate data points yields the tester's complexity fitness. Taking tester E1 as an example, their complexity fitness is 0.75, representing their adaptability to the current task complexity level.

[0059] The average complexity of all tasks within a task cluster is calculated by summing the complexities of each task within the cluster and dividing by the number of tasks. For example, if task cluster F1 contains 3 tasks with complexities of 0.6, 0.7, and 0.8 respectively, then the average complexity of this task cluster is (0.6 + 0.7 + 0.8) / 3 = 0.7. Assume there are 4 task clusters with the following average complexities: F1 = 0.7, F2 = 0.85, F3 = 0.65, and F4 = 0.78.

[0060] When calculating the matching degree between inspectors and task clusters, the influence of task carrying capacity, collaborative affinity, and task complexity needs to be considered. A first balance coefficient of 0.4, a second balance coefficient of 0.3, and a third balance coefficient of 0.3 are set. Collaborative affinity, obtained through the first-level clustering, represents the degree of affinity between inspectors and task clusters. For example, the collaborative affinity between inspector E1 and task cluster F1 is 0.82. Complexity fitness represents the inspector's ability to adapt to tasks of specific complexity and can be obtained through historical data analysis. Assume that inspector E1's complexity fitness is 0.75.

[0061] The matching degree calculation process for the testing personnel E1 and task cluster F1 is as follows: multiply E1's task carrying capacity index 0.85 by the first balance coefficient 0.4 to get 0.34, multiply the cooperative affinity 0.82 by the second balance coefficient 0.3 to get 0.246, calculate the difference between the average complexity of task cluster F1 0.7 and the complexity fitness of E1 0.75, which is -0.05, multiply this difference by the third balance coefficient 0.3 to get -0.015, and finally add the above results to get the matching degree as 0.34 + 0.246 + 0.015 = 0.601.

[0062] Through similar calculations, the matching degree matrix of all inspectors and task clusters can be obtained. Taking 5 inspectors and 4 task clusters as an example, the matching degree matrix is ​​as follows: the matching degree of inspector E1 to task clusters F1 to 4 is 0.601, 0.534, 0.625, and 0.569, respectively; the matching degree of inspector E2 to task clusters F1 to 4 is 0.642, 0.687, 0.598, and 0.615, respectively; the matching degree of inspector E3 to task clusters F1 to 4 is 0.521, 0.492, 0.568, and 0.507, respectively; the matching degree of inspector E4 to task clusters F1 to 4 is 0.633, 0.659, 0.612, and 0.645, respectively; and the matching degree of inspector E5 to task clusters F1 to 4 is 0.547, 0.523, 0.581, and 0.536, respectively.

[0063] Based on the matching degree matrix, a genetic algorithm is used to optimize task cluster allocation. The chromosome encoding of the genetic algorithm adopts a multi-genomic structure, with each gene locus corresponding to a task cluster. The value of the gene locus is the ID of the tester assigned to that task cluster. For example, chromosome [E2,E4,E1,E3] indicates that the task clusters are assigned to testers E2, E4, E1, and E3, respectively.

[0064] The fitness function is defined as the sum of the matching degrees of all inspectors and their assigned task clusters, minus the constraint violation penalty. The constraints include: each inspector's workload cannot exceed their capacity limit; each task cluster can only be assigned to one inspector. If an inspector's workload exceeds the limit, a penalty of 20% of the excess is calculated.

[0065] The initial population randomly generates multiple chromosomes, for example, an initial population of size 50. The fitness value of each chromosome is calculated and sorted in descending order of fitness value. Taking the matching matrix above as an example, assuming a randomly generated chromosome is [E2, E4, E1, E3], its fitness value is calculated as follows: Matching degree between inspector E2 and task cluster F1: 0.642 + Matching degree between inspector E4 and task cluster F2: 0.659 + Matching degree between inspector E1 and task cluster F3: 0.625 + Matching degree between inspector E3 and task cluster F4: 0.507 = 2.433.

[0066] The crossover operation in the genetic algorithm uses partial mapping crossover. Two parent chromosomes are selected from the population, and a crossover region is randomly determined. The values ​​of gene positions are exchanged within this region. For example, if parent chromosomes P1=[E2,E4,E1,E3] and P2=[E1,E3,E4,E2], and the crossover region is the 2nd and 3rd positions, the resulting offspring chromosomes are R1=[E2,E3,E4,E3] and R2=[E1,E4,E1,E2]. Since the detector E3 appears twice in R1 and the detector E1 appears twice in R2, mapping repair is required. Based on the mapping relationships E4->E3 and E1->E4 in the crossover region, the 4th position E3 in R1 is changed to E1, resulting in R1=[E2,E3,E4,E1]; the 3rd position E1 in R2 is changed to E3, resulting in R2=[E1,E4,E3,E2].

[0067] The mutation operation involves randomly selecting two gene loci in a chromosome and exchanging their corresponding tester numbers. For example, to mutate chromosome [E2,E3,E4,E1], the 1st and 4th loci are randomly selected and exchanged to obtain the mutated chromosome [E1,E3,E4,E2].

[0068] The selection, crossover, and mutation operations are repeated until the maximum number of iterations or the fitness value converges. Assuming a maximum number of iterations of 100, after iterations, the chromosome with the highest fitness value [E3, E2, E5, E1] is obtained, indicating that task cluster F1 is assigned to tester E3, task cluster F2 to tester E2, task cluster F3 to tester E5, and task cluster F4 to tester E1. This scheme has a fitness value of 2.675, higher than the initial random scheme, indicating that the matching degree between testers and task clusters is significantly improved after optimization using the genetic algorithm.

[0069] In one optional implementation, chromosomes are sorted according to their fitness values ​​from largest to smallest. Chromosomes are selected according to this sorting and subjected to partial mapping crossover. Gene loci values ​​are exchanged within the crossover region. The mutation operation is completed by randomly selecting two gene loci from the chromosomes and exchanging their corresponding personnel numbers. Chromosomes are sorted from largest to smallest fitness value. The ratio of the current chromosome's fitness value to the population's optimal fitness value is calculated based on the fitness value to obtain the population selection ratio. Parent chromosomes and offspring chromosomes are selected based on the sorting and the population selection ratio. Two chromosomes are randomly selected from the chosen parent chromosomes for partial mapping crossover. The basic crossover coefficient is summed with the negative exponential function of the current iteration number and multiplied by the chromosome length to obtain the crossover region length. The one-to-one correspondence between the gene positions of the two parent chromosomes is recorded in the crossover region to form a mapping table. The detection personnel number of the second parent chromosome in the crossover region is copied to the corresponding position of the offspring chromosome. The duplicate detection personnel number in the offspring chromosome is eliminated according to the mapping table, and the exchange of gene position values ​​is completed. The task contribution of each gene locus in the offspring chromosome is calculated by summing the matching degree between the task cluster corresponding to each gene locus and the assigned testing personnel. The task contribution of each gene locus is subtracted from the maximum task contribution of all gene loci and normalized to obtain the mutation selection probability. Two mutation positions are determined according to the mutation selection probability. The testing personnel numbers corresponding to the two mutation positions are swapped to complete the mutation operation.

[0070] In this embodiment, a chromosome represents a detection task allocation scheme, where each gene locus corresponds to a task cluster, and the value of the gene locus is the detection personnel number assigned to that task cluster. First, the fitness value of each chromosome is calculated. This fitness value reflects the quality of the allocation scheme; a higher fitness value indicates a better allocation scheme.

[0071] All chromosomes are sorted in descending order of fitness value. The best fitness value is denoted as `best_fitness`, and the fitness value of the current chromosome is denoted as `current_fitness`. The selection ratio is calculated as `selection_ratio = current_fitness / best_fitness`. This ratio determines the selection range for parent and offspring chromosomes. For example, in a population of 100 chromosomes, if a chromosome has a selection_ratio of 0.8, then the parent chromosome can be selected from the top 80 chromosomes, and the offspring chromosome to be replaced can be selected from the bottom 20 chromosomes.

[0072] When selecting parent chromosomes, a roulette wheel method is used, where chromosomes with higher fitness values ​​have a greater probability of being selected. Assuming two parent chromosomes, P3 and P4, are selected, with P3 having the gene sequence [E5, E2, E8, E4, E7, E1, E3, E6] and P4 having the gene sequence [E3, E7, E5, E1, E6, E8, E2, E4], this means that eight task clusters are assigned to the detection personnel whose numbers correspond to the values.

[0073] Perform a partial crossover operation on the selected parent chromosomes. First, determine the crossover region length. Set the crossover coefficient `base_crossover_rate` to 0.3, the current iteration count to 50, and the total iteration count to 200. Then, calculate the crossover region length as (base_crossover_rate + exp(-iteration / 200)) × chromosome length. Taking a chromosome length of 8 as an example, the calculated crossover region length is approximately 3. Randomly select the starting position for crossover; assuming it starts from position 2, the crossover regions will be positions E2, E3, and E4.

[0074] Within the crossover region, the correspondence between the gene loci of the two parent chromosomes is recorded, forming a mapping table: E2→E7, E8→E5, E4→E1. The gene value of P4 within the crossover region is copied to the corresponding position on the offspring chromosome, resulting in the preliminary offspring sequence [?, E7, E5, E1, ?, ?, ?, ?], where ? represents the position to be determined.

[0075] According to the mapping table, duplicates are eliminated. The gene values ​​in P3 are traversed. If a gene value is not in the crossover region and has not appeared in the offspring, it is directly copied to the corresponding position in the offspring. If a gene value has already appeared in the offspring, it is replaced according to the mapping table until an unused gene value is found. After processing according to this rule, the complete offspring chromosome offspring is obtained as [E5, E7, E5, E1, E6, E8, E2, E4].

[0076] Next, the mutation operation is performed. First, the task contribution of each gene locus in the offspring chromosome is calculated. Assume there is a pre-calculated matching matrix match_matrix between task clusters and testers, where match_matrix[i][j] represents the matching degree of the i-th task cluster assigned to the j-th tester. For the offspring chromosome offspring, the task contribution of each gene locus i is calculated as contribution[i] = match_matrix[i][offspring[i]].

[0077] For example, in the case of 8 task clusters and 8 inspectors, the matching degree matrix partial values ​​are as follows: the matching degree between task cluster 0 and inspector 5 is 0.75, the matching degree between task cluster F1 and inspector E6 is 0.62, and so on. The calculated task contribution values ​​for each gene locus are [0.75, 0.62, 0.58, 0.81, 0.70, 0.65, 0.77, 0.69].

[0078] Find the maximum task contribution, max_contribution = 0.81, and calculate the mutation probabilities for each gene locus: mutation_prob[i] = (max_contribution - contribution[i]) / sum(max_contribution - contribution). After normalization, the selection probabilities for each position are [0.13, 0.22, 0.27, 0.0, 0.13, 0.18, 0.05, 0.14].

[0079] Based on the above probabilities, two mutation positions are selected using the roulette wheel method. Suppose positions 2 and 5 are selected, which are the positions with gene values ​​of E7 and E8. The tester numbers for these two positions are swapped, and the offspring chromosome offspring_mutated after mutation is obtained as [E5, E8, E5, E1, E6, E7, E2, E4].

[0080] After mutation, the newly generated offspring chromosomes replace the chromosomes with lower fitness in the population. Repeat the selection, crossover, and mutation operations until a sufficient number of offspring chromosomes are generated. Through multiple generations of iteration, the chromosomes in the population will gradually converge towards the optimal solution, ultimately yielding the optimal task allocation scheme.

[0081] Experiments show that, for the large-scale allocation problem of 100 task clusters and 20 inspectors, the allocation scheme obtained by this method after 300 iterations improves the task completion efficiency by 15% compared with the traditional genetic algorithm, while reducing the workload imbalance of inspectors by 23%, proving the effectiveness of the method of this invention in complex task allocation scenarios.

[0082] This invention relates to a big data-based intelligent allocation system for laboratory testing tasks, the system comprising: The first unit is used to collect data for the detection task; The second unit is used to perform a decomposability analysis on the detection task, decompose the task into multiple collaborative detection sub-tasks, construct a task collaboration flowchart based on the dependencies between sub-tasks, determine the execution order and data interaction nodes, and calculate the collaborative efficiency index of the detection team based on the task collaboration flowchart. The third unit is used to calculate the task allocation priority score based on the detection task data and the collaborative efficiency index, and generate a task queue to be allocated. The fourth unit is used to extract the collaboration pattern features of the inspectors from historical collaborative inspection records and generate a correlation weight matrix between the inspectors, and calculate the collaboration affinity between the inspectors based on the correlation weight matrix. The fifth unit is used to allocate the task queue based on a two-layer adaptive task allocation algorithm. In the first layer, task feature vectors are constructed, and the cosine similarity method is used to calculate the similarity matrix between tasks. Density peak clustering is performed based on the similarity matrix, and cluster centers are identified by local density and distance between tasks, dividing the tasks to be allocated into multiple task clusters. In the second layer, the task carrying capacity index of the detection personnel is obtained, and the matching degree matrix between the task clusters and the detection personnel is calculated by combining the cooperative affinity. The allocation scheme of each task cluster is optimized by using a genetic algorithm, and the final scheme is obtained through iteration. The sixth unit is used to distribute the final solution to the testing personnel's terminal.

[0083] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0084] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0085] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent allocation of laboratory testing tasks based on big data, characterized in that, include: Collect detection task data; The detection task is decomposed into multiple collaborative detection subtasks by performing a decomposability analysis. A task collaboration flowchart is constructed based on the dependencies between the subtasks to determine the execution order and data interaction nodes. Based on the task collaboration flowchart, the collaborative efficiency index of the detection team is calculated. Calculate the task allocation priority score based on the detection task data and the collaborative efficiency index, and generate a task queue to be allocated. Extract the collaboration pattern features of the inspectors from historical collaborative inspection records and generate a correlation weight matrix between the inspectors. Calculate the collaboration affinity between the inspectors based on the correlation weight matrix. The task queue is allocated based on a two-layer adaptive task allocation algorithm. In the first layer, task feature vectors are constructed, and the cosine similarity method is used to calculate the similarity matrix between tasks. Density peak clustering is performed based on the similarity matrix, and cluster centers are identified by local density and distance between tasks, dividing the tasks to be allocated into multiple task clusters. In the second layer, the task carrying capacity index of the detection personnel is obtained, and the matching degree matrix between the task clusters and the detection personnel is calculated by combining the cooperative affinity. The allocation scheme of each task cluster is optimized by a genetic algorithm, and the final scheme is obtained through iteration. The final solution is then sent to the testing personnel's terminal.

2. The method according to claim 1, characterized in that, The detection task is decomposed into multiple collaborative detection subtasks based on a decomposability analysis. A task collaboration flowchart is constructed based on the dependencies between the subtasks to determine the execution order and data interaction nodes. Based on this task collaboration flowchart, the collaborative efficiency index of the detection team is calculated, including: The detection task is recursively decomposed into multiple collaborative detection subtasks. Each collaborative detection subtask is marked with a task flow node, and a task flow matrix is ​​generated. Based on the task flow matrix, the execution timing constraints of the collaborative detection subtasks are analyzed, and directed connections are established between collaborative detection subtasks with pre-determined dependencies to generate a task collaboration flowchart. Using collaborative detection subtasks as vertices, data transfer relationships between collaborative detection subtasks as edges, and data interaction volume as edge weights, the vertex degree and betweenness centrality are calculated; the critical path is determined based on the directed connection relationships in the task collaboration flowchart, and the importance of the data interaction node is calculated based on the weights of the critical path, the vertex degree, and the betweenness centrality. The collaborative execution efficiency is obtained by extracting the ratio of collaborative execution time to independent execution time from historical detection records, and the degree of parallel execution of collaborative detection subtasks is calculated based on the topological structure of the task collaboration flowchart. The collaborative efficiency index of the detection team is obtained by multiplying the importance of the data interaction node, the collaborative execution efficiency, and the degree of parallel execution.

3. The method according to claim 1, characterized in that, The collaboration pattern features of the inspectors are extracted from historical collaborative inspection records, and a correlation weight matrix between the inspectors is generated. The collaboration affinity between the inspectors is calculated based on the correlation weight matrix, including: The collaboration duration feature, collaboration frequency feature, and quality score feature are extracted from historical collaborative detection records and normalized by maximum and minimum values ​​to obtain normalized feature values. Multiply the normalized average collaboration time by the duration weight coefficient, the normalized collaboration density by the frequency weight coefficient, and the normalized average quality score by the quality weight coefficient to obtain the multidimensional feature weight matrix. Calculate the L2 norm of the multidimensional feature weight matrix for each pair of inspectors, and divide the L2 norm by the maximum value of the L2 norm among all pairs of inspectors to obtain the collaboration strength coefficient; extract the quality fluctuation coefficient of each pair of inspectors and normalize it, and subtract the normalized quality fluctuation coefficient from 1 to obtain the collaboration stability coefficient; count the historical collaboration times of each pair of inspectors, and divide the collaboration times by the maximum value of the collaboration times among all pairs of inspectors to obtain the collaboration experience coefficient. The degree of synergy between the testing personnel is obtained by multiplying the synergy strength coefficient by the first comprehensive weighting coefficient, the synergy stability coefficient by the second comprehensive weighting coefficient, and the synergy experience coefficient by the third comprehensive weighting coefficient, and then summing the three weighted coefficients.

4. The method according to claim 1, characterized in that, In the first layer, task feature vectors are constructed, and the cosine similarity method is used to calculate the inter-task similarity matrix. Density peak clustering is performed based on the similarity matrix, and cluster centers are identified through local density and inter-task distance, dividing the tasks to be assigned into multiple task clusters, including: Extract the feature parameters of the detection task and multiply them by the corresponding feature weight coefficients to obtain the task feature vector; Calculate the inner product and norm product of the feature vectors of two detection tasks, divide the inner product by the norm product to obtain the similarity between the two detection tasks, and traverse all detection task pairs to generate a task similarity matrix. Subtract the similarity in the task similarity matrix from 1 to obtain the task distance matrix; select the maximum distance value in the task distance matrix, and multiply the maximum distance value by the density truncation coefficient to obtain the truncation distance; count the number of tasks whose distance from each detection task to other detection tasks is less than the truncation distance, and obtain the local density of the detection task. Calculate the distance between each detection task and the first detection task, and select the smallest distance value as the minimum distance of the first detection task; calculate the distance between the detection task with the highest local density and all other detection tasks, and select the largest distance value as the minimum distance of that detection task; repeat the above process for each detection task to obtain its minimum distance, and multiply the local density of each detection task by its corresponding minimum distance to obtain the decision value of that detection task. The local maximum value among the decision values ​​is selected as the cluster center, and the remaining detection tasks are assigned to the nearest cluster center to form multiple task clusters.

5. The method according to claim 1, characterized in that, In the second layer, the task capacity index of the testing personnel is obtained, and the matching degree matrix between task clusters and testing personnel is calculated by combining the cooperative affinity. The allocation scheme of each task cluster is optimized using a genetic algorithm, and the final scheme is obtained through iteration, including: The task carrying capacity index is obtained based on the historical tasks of the testing personnel; Extract the success rate of inspectors completing tasks of different complexities from historical inspection records, and fit the success rate with the corresponding task complexity to obtain the complexity fitness. The average complexity of all tasks within a task cluster is calculated to obtain the cluster-wide task complexity. The task carrying capacity index is multiplied by a first balance coefficient, and the cooperative affinity is multiplied by a second balance coefficient. The first and second balance coefficients are added together, and the difference between the cluster-wide task complexity and the complexity fitness is subtracted and multiplied by a third balance coefficient to obtain the matching degree matrix between the testing personnel and the task cluster. The fitness value is obtained by calculating the sum of the matching degrees of all inspectors and their assigned task clusters based on the matching degree matrix and subtracting the constraint violation penalty term. Construct a multi-genomic chromosome, with each gene locus corresponding to a task cluster. The gene locus value is assigned to the detection personnel number of that task cluster. Sort the chromosomes according to their fitness values ​​from largest to smallest. Select chromosomes according to the sorting and perform partial mapping crossover operations, exchanging the values ​​of gene loci within the crossover region. Randomly select two gene loci in the chromosomes and exchange their corresponding detection personnel numbers to complete the mutation operation. Repeat the crossover operation and the mutation operation until the maximum number of iterations is reached. Select the chromosome code corresponding to the highest fitness value as the final task allocation scheme.

6. The method according to claim 5, characterized in that, Chromosomes are sorted from highest to lowest fitness value. Chromosomes are selected based on this sorting and subjected to partial mapping crossover. Gene loci values ​​are exchanged within the crossover region. Finally, two gene loci on randomly selected chromosomes are swapped with their corresponding personnel IDs to complete the mutation operation. Chromosomes are sorted from largest to smallest fitness value. The ratio of the current chromosome's fitness value to the population's optimal fitness value is calculated based on the fitness value to obtain the population selection ratio. Parent chromosomes and offspring chromosomes are selected based on the sorting and the population selection ratio. Two chromosomes are randomly selected from the chosen parent chromosomes for partial mapping crossover. The basic crossover coefficient is summed with the negative exponential function of the current iteration number and multiplied by the chromosome length to obtain the crossover region length. The one-to-one correspondence between the gene positions of the two parent chromosomes is recorded in the crossover region to form a mapping table. The detection personnel number of the second parent chromosome in the crossover region is copied to the corresponding position of the offspring chromosome. The duplicate detection personnel number in the offspring chromosome is eliminated according to the mapping table, and the exchange of gene position values ​​is completed. The task contribution of each gene locus in the offspring chromosome is calculated by summing the matching degree between the task cluster corresponding to each gene locus and the assigned testing personnel. The task contribution of each gene locus is subtracted from the maximum task contribution of all gene loci and normalized to obtain the mutation selection probability. Two mutation positions are determined according to the mutation selection probability. The testing personnel numbers corresponding to the two mutation positions are swapped to complete the mutation operation.

7. A big data-based intelligent allocation system for laboratory testing tasks, used to implement the method as described in any one of claims 1-6, characterized in that, include: The first unit is used to collect data for the detection task; The second unit is used to perform a decomposability analysis on the detection task, decompose the task into multiple collaborative detection sub-tasks, construct a task collaboration flowchart based on the dependencies between sub-tasks, determine the execution order and data interaction nodes, and calculate the collaborative efficiency index of the detection team based on the task collaboration flowchart. The third unit is used to calculate the task allocation priority score based on the detection task data and the collaborative efficiency index, and generate a task queue to be allocated. The fourth unit is used to extract the collaboration pattern features of the inspectors from historical collaborative inspection records and generate a correlation weight matrix between the inspectors, and calculate the collaboration affinity between the inspectors based on the correlation weight matrix. The fifth unit is used to allocate the task queue based on a two-layer adaptive task allocation algorithm. In the first layer, task feature vectors are constructed, and the cosine similarity method is used to calculate the similarity matrix between tasks. Density peak clustering is performed based on the similarity matrix, and cluster centers are identified by local density and distance between tasks, dividing the tasks to be allocated into multiple task clusters. In the second layer, the task carrying capacity index of the detection personnel is obtained, and the matching degree matrix between the task clusters and the detection personnel is calculated by combining the cooperative affinity. The allocation scheme of each task cluster is optimized by using a genetic algorithm, and the final scheme is obtained through iteration. The sixth unit is used to distribute the final solution to the testing personnel's terminal.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images