Method and system for evaluating matching degree between crew and post based on multi-dimensional feature weighting

Abstract

The present application relates to a kind of crew and post matching degree evaluation method and system based on multi-dimensional feature weighting, method includes: obtaining the multi-dimensional feature data of crew and ship owner, generates crew initial feature vector and ship owner initial feature vector;Based on historical evaluation data, the evaluation score of multiple parties is weighted and fused and moving average, generate crew initial credit score and ship owner initial credit score;Build the evaluation relationship bipartite graph of crew node and ship owner node, calculate global coupling strength to quantify the degree of interference of two-way evaluation;Crew credit score and ship owner credit score with significant coupling are decoupled and corrected by covariance, eliminate the coupling distortion in two-way credit evaluation and generate decoupled credit score;Based on decoupled credit score, generate post demand feature vector after credit threshold verification to post demand, carry out dimension-by-dimension deviation calculation and self-adapting weighted summation to crew multi-dimensional feature weighted vector, generate weighted comprehensive matching degree, to indicate the adaptation degree of crew and current post.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for evaluating the matching degree between crew members and positions based on multi-dimensional feature weighting. Background Technology

[0002] With the continuous improvement of informatization in the shipping industry, data-driven crew job matching assessment methods have gradually become a research hotspot. These methods typically collect multi-dimensional characteristic data such as crew member certifications, service experience, and safety records to construct quantitative assessment models, assisting shipowners in screening crew members who match job requirements during the recruitment process. Compared to traditional recruitment methods that rely on manual resume screening and interviews, crew job matching methods based on multi-dimensional characteristic assessment can improve recruitment efficiency and matching accuracy to a certain extent, and therefore have received widespread attention from shipping companies and relevant management departments.

[0003] However, existing methods for assessing crew job matching still have significant technical shortcomings. On the one hand, most existing solutions only evaluate crew members from the perspective of the shipowner or management, lacking a mechanism for crew members to assess the creditworthiness of the shipowner, thus failing to provide crew members with an equivalent decision-making reference when choosing an employer. On the other hand, even in scenarios involving two-way evaluation, existing methods typically calculate the credit scores of both parties independently, ignoring the potential interference between the shipowner's credit level and the crew member's evaluation. When a crew member serves a shipowner with abnormal creditworthiness for a long period, their own credit score may be unreasonably inflated or deflated, and vice versa. This coupling effect prevents credit scores from truly reflecting the independent responsibility of the evaluating party, leading to systematic biases in job matching results based on credit scores. Furthermore, existing methods lack a refined technical connection between credit scores and job matching; the weights of each dimension during the matching process are often fixed and cannot be adaptively adjusted according to job type and the specific employment characteristics of the shipowner. Summary of the Invention

[0004] Based on this, the purpose of this invention is to provide a method and system for evaluating the matching degree between crew members and positions based on multi-dimensional feature weighting, which can eliminate the coupling interference of two-way credit evaluation, realize the independent decoupling of credit scores, and perform adaptive matching of positions accordingly.

[0005] The objective of this invention is achieved through the following solution:

[0006] In a first aspect, the present invention provides a method for evaluating the matching degree between crew members and positions based on multi-dimensional feature weighting, comprising the following steps:

[0007] S1: Obtain multi-dimensional feature data of crew members and ship owners, perform vectorization encoding on the multi-dimensional feature data, and generate initial feature vectors for crew members and ship owners;

[0008] S2: Based on the historical evaluation data in the multidimensional feature data, the multi-party evaluation scores in the historical evaluation data are weighted and summed to calculate the comprehensive score of a crew member's single contract. The comprehensive score of a crew member's single contract is then weighted and moved averaged to generate the initial credit score of the crew member and the initial credit score of the shipowner.

[0009] S3: Construct a bipartite graph of the evaluation relationship between crew members and ship owners based on historical evaluation data. Perform deviation product calculation on the evaluation score on each edge of the bipartite graph and the initial credit scores of the crew members and ship owners at the corresponding two ends of the node to generate the global coupling strength between each crew member node and each ship owner node.

[0010] S4: Based on the global coupling strength, select the crew credit scores and shipowner credit scores that have significant coupling, and perform covariance decoupling correction on the crew credit scores and shipowner credit scores that have significant coupling to generate crew decoupled credit scores and shipowner decoupled credit scores.

[0011] S5: Based on the shipowner's decoupled credit score, the credit threshold verification is performed on the job requirement data obtained from the shipowner's terminal. The minimum credit score requirement for the job requirement is compared with the shipowner's decoupled credit score. Invalid job requirements that do not meet the credit threshold are eliminated, and the job requirement feature vector is output. The initial feature vector of the crew member is concatenated with the crew member's decoupled credit score to generate a multi-dimensional feature weighted vector of the crew member. The multi-dimensional feature weighted vector of the crew member is combined with the job requirement feature vector to perform a dimension-by-dimensional deviation calculation and weighted summation to generate a weighted comprehensive matching degree between the crew member and the job.

[0012] The weighted overall matching degree is used to indicate the degree of suitability between the crew member and the current position. The higher the matching degree, the more suitable the crew member is for the current position.

[0013] Secondly, the present invention provides a crew member and job matching evaluation system based on multi-dimensional feature weighting, which is configured with the following modules:

[0014] The multidimensional feature encoding module is used to acquire multidimensional feature data of crew members and ship owners, and to perform vectorization encoding on the multidimensional feature data to generate initial feature vectors for crew members and ship owners.

[0015] The initial credit calculation module is used to calculate the comprehensive score of a single contract for crew members based on the historical evaluation data in the multidimensional feature data, by weighting and summing the multi-party evaluation scores in the historical evaluation data, and by weighting and moving average the comprehensive score of a single contract for crew members to generate the initial credit score of crew members and the initial credit score of ship owners.

[0016] The bipartite graph coupling strength solution module is used to construct a bipartite graph of the evaluation relationship between crew members and ship owners based on historical evaluation data. It performs deviation product calculation on the evaluation score on each edge of the bipartite graph and the initial credit scores of the crew members and ship owners at the corresponding two ends of the node to generate the global coupling strength between each crew member node and each ship owner node.

[0017] The credit score covariance decoupling module is used to filter out crew credit scores and shipowner credit scores with significant coupling based on the global coupling strength, and to perform covariance decoupling correction processing on the crew credit scores and shipowner credit scores with significant coupling, thereby generating crew decoupled credit scores and shipowner decoupled credit scores.

[0018] The comprehensive matching calculation module for personnel and positions is used to perform credit threshold verification on the job requirement data obtained from the shipowner's terminal based on the shipowner's decoupled credit score. It compares the minimum credit score requirement of the job requirement with the shipowner's decoupled credit score, eliminates invalid job requirements that do not meet the credit threshold, and outputs the job requirement feature vector. It concatenates the initial feature vector of the crew member with the crew member's decoupled credit score to generate a multi-dimensional feature weighted vector of the crew member. It then calculates the deviation of the multi-dimensional feature weighted vector of the crew member by dimension and performs weighted summation based on the job requirement feature vector to generate a weighted comprehensive matching degree between the crew member and the job.

[0019] The weighted overall matching degree is used to indicate the degree of suitability between the crew member and the current position. The higher the matching degree, the more suitable the crew member is for the current position.

[0020] Thirdly, this application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the above-mentioned methods for evaluating the matching degree between crew members and positions based on multi-dimensional features.

[0021] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-mentioned methods for evaluating the matching degree between crew members and positions based on multi-dimensional feature weighting.

[0022] In summary, the multi-dimensional feature-weighted crew-job matching assessment method provided in this application effectively eliminates the coupling interference between shipowner credit levels and crew evaluation scores by constructing a two-way symmetrical credit evaluation framework for both crew members and shipowners and performing covariance decoupling correction on the credit scores of both ends. This ensures that the credit score truly reflects the independent performance responsibility of the evaluation subject, thereby achieving equivalence and fairness in credit evaluation between crew members and shipowners. Simultaneously, this method integrates the decoupled crew credit score as an independent dimension into the multi-dimensional feature vector and adaptively assigns weights to each feature dimension based on job type characteristics and the shipowner's historical operating profile. This allows for dynamic adjustment of the matching focus according to the actual needs of different positions, significantly improving the accuracy of person-job matching. Furthermore, this method pre-verifies the credit threshold for job requirements by decoupling the shipowner's credit score, eliminating invalid positions with unqualified credit from the outset. It also generates a unified matching metric based on dimensional deviation calculation and weighted summation, enabling seamless data flow from credit decoupling to job matching, providing objective and quantifiable technical support for crew recruitment decisions.

[0023] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0024] Figure 1 A flowchart illustrating a method for evaluating crew member-job matching based on multi-dimensional feature weighting, provided in an embodiment of this application;

[0025] Figure 2 This is a schematic diagram of the structure of a crew member and job matching evaluation system based on multi-dimensional feature weighting, which is provided as another embodiment of this application. Detailed Implementation

[0026] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0028] In one embodiment, such as Figure 1As shown, a method for evaluating the matching degree between crew members and positions based on multi-dimensional feature weighting is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and can also be applied to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0029] S1: Obtain multidimensional feature data of crew members and ship owners, perform vectorization encoding on the multidimensional feature data, and generate initial feature vectors for crew members and ship owners.

[0030] Specifically, for seafarers, the system collects information on certificate level, certificate type, certificate validity period, type of service vessel, tonnage of service vessel, service route, service duration, safety records, training records, contract performance, and platform activity. For shipowners, the system collects business registration information, operating qualification information, ship management records, historical recruitment information, wage payment records, contract execution records, and complaint handling records. The system performs vectorization encoding on the acquired multidimensional feature data. During vectorization encoding, discrete feature data is converted into numerical vectors using one-hot encoding or dictionary mapping encoding, and continuous feature data is mapped to a unified numerical range using normalization or standardization. The system concatenates all the encoded feature vectors according to a preset dimensional order to generate initial feature vectors for seafarers and shipowners. The initial feature vector for seafarers is used to characterize the quantitative attributes of seafarers in terms of professional skills, safety awareness, professional ethics, and credit behavior. The initial feature vector for shipowners is used to characterize the quantitative attributes of shipowners in terms of compensation and benefits, work environment, communication and respect, and platform cooperation.

[0031] S2: Based on historical evaluation data in multidimensional feature data, the multi-party evaluation scores in the historical evaluation data are weighted and summed to calculate the comprehensive score of a crew member's single contract. The comprehensive score of a crew member's single contract is then weighted and moved averaged to generate the initial credit score of the crew member and the initial credit score of the shipowner.

[0032] Specifically, the system extracts the shipowner evaluation scores, anonymous peer evaluation scores, and platform objective behavior evaluation scores received by crew members during each contract period. The system assigns preset weights to each type of evaluation source and weights and sums the evaluation scores from different sources within the same contract period to obtain the crew member's comprehensive score for a single contract. For shipowners, the system extracts the crew evaluation scores and platform objective data scores received during each contract period, and weights and sums them according to preset weights for crew and platform evaluations to obtain the shipowner's comprehensive score for a single contract. The system obtains a complete sequence of historical comprehensive scores for all crew members or shipowners' single contracts, arranges them in chronological order of contract end time, and processes the score sequence using a weighted moving average algorithm. In the weighted moving average processing, the system assigns higher weight coefficients to contracts more recent than the current time and lower weight coefficients to contracts more distant than the current time. The system generates initial credit scores for crew members and shipowners based on the weighted moving average calculation results. The initial credit score of the crew and the initial credit score of the shipowner each represent a quantitative value of the overall credit level of the corresponding entity throughout all historical contract periods.

[0033] S3: Construct a bipartite graph of the evaluation relationship between crew members and ship owners based on historical evaluation data. Perform deviation product calculation on the evaluation score of each edge in the bipartite graph with the initial credit scores of the crew members and ship owners at the corresponding two ends of the node to generate the global coupling strength between each crew member node and each ship owner node.

[0034] Specifically, the system constructs a bipartite graph of evaluation relationships between crew members and shipowners based on historical evaluation data. Each crew member is designated as a first-type node, and each shipowner as a second-type node. The system iterates through each evaluation record in the historical data. When an evaluation record indicates that a crew member has evaluated a shipowner or a shipowner has evaluated a crew member, the system establishes an edge between that crew member node and that shipowner node. The evaluation score of that evaluation is used as the initial weight value of the edge. For each edge, the system obtains the initial credit score of the crew member node and the initial credit score of the shipowner node connected to the edge, calculates the first difference between the evaluation score on the edge and the initial credit score of the crew member, and calculates the second difference between the evaluation score on the edge and the initial credit score of the shipowner. The system multiplies the first and second differences to obtain the deviation product value of the edge. The system calculates the global coupling strength between a crew member node and a shipowner node by averaging the product of the deviations of multiple edges connecting the same crew member node and the same shipowner node, after summing the products and dividing by the number of edges. A positive global coupling strength indicates a positive coupling relationship between the crew member's credit score and the shipowner's credit score, while a negative global coupling strength indicates a negative coupling relationship.

[0035] S4: Based on the global coupling strength, select the crew credit scores and shipowner credit scores that have significant coupling. Perform covariance decoupling correction on the crew credit scores and shipowner credit scores that have significant coupling to generate crew decoupled credit scores and shipowner decoupled credit scores.

[0036] Specifically, the system statistically analyzes the global coupling strength between all crew member nodes and shipowner nodes, calculating the distribution parameters of the global coupling strength. The system determines a coupling strength threshold based on these distribution parameters. The system classifies node pairs with an absolute global coupling strength greater than the coupling strength threshold as having significant coupling between their corresponding crew member credit scores and shipowner credit scores. The system then performs covariance decoupling correction processing on the significantly coupled crew member credit scores and shipowner credit scores. In this process, the system constructs a two-dimensional variable set using the initial crew member credit score and the initial shipowner credit score, and calculates the covariance matrix of this set. The system determines the direction and degree of linear correlation between the crew member credit score and the shipowner credit score based on the covariance elements in the covariance matrix. The system constructs a decoupling transformation matrix to map the original two-dimensional variable set to a new coordinate space where the covariance between the crew member dimension and the shipowner dimension is zero. The system inputs the significantly coupled crew member credit scores and shipowner credit scores into the decoupling transformation matrix for linear transformation, generating decoupled crew member credit scores and decoupled shipowner credit scores. For crew credit scores and shipowner credit scores that are not significantly coupled, the system directly outputs their initial credit scores as decoupled credit scores.

[0037] S5: Based on the shipowner's decoupled credit score, the credit threshold verification is performed on the job requirement data obtained from the shipowner's terminal. The minimum credit score requirement for the job requirement is compared with the shipowner's decoupled credit score. Invalid job requirements that do not meet the credit threshold are eliminated, and the job requirement feature vector is output. The initial feature vector of the crew member is concatenated with the crew member's decoupled credit score to generate a multi-dimensional feature weighted vector of the crew member. The multi-dimensional feature weighted vector of the crew member is combined with the job requirement feature vector to perform a dimension-by-dimensional deviation calculation and weighted summation to generate a weighted comprehensive matching degree between the crew member and the job.

[0038] The weighted overall matching degree is used to indicate the degree of suitability between the crew member and the current position. The higher the matching degree, the more suitable the crew member is for the current position.

[0039] Specifically, the system extracts the minimum credit score threshold required for the position from the job requirement data and compares the shipowner's decoupling credit score with the minimum credit score threshold. When the shipowner's decoupling credit score is greater than or equal to the minimum credit score threshold, the system determines the job requirement to be valid. When the shipowner's decoupling credit score is less than the minimum credit score threshold, the system determines the job requirement to be invalid and removes it from the subsequent matching process. The system vectorizes the valid job requirements that pass the credit threshold verification, generating a job requirement feature vector. The job requirement feature vector contains quantified values ​​for dimensions such as ship type requirements, tonnage requirements, route requirements, certificate level requirements, and skill requirements. The system obtains the crew member's initial feature vector and crew member's decoupling credit score, concatenating the crew member's decoupling credit score as an independent dimension to the end of the crew member's initial feature vector to generate a crew member multi-dimensional feature weighted vector. For each dimension in the job requirement feature vector, the system calculates the absolute value of the deviation between the value of that dimension and the corresponding dimension value in the crew member multi-dimensional feature weighted vector. The system assigns a preset weight coefficient to each dimension of the job requirement feature vector. The absolute value of the deviation for each dimension is multiplied by its corresponding weight coefficient, and then summed to obtain a weighted sum of deviations between the crew member and the job. The system converts this weighted sum of deviations into a matching score according to a preset mapping relationship, generating a weighted comprehensive matching degree between the crew member and the job. This weighted comprehensive matching degree indicates the degree of fit between the crew member and the current job.

[0040] In summary, the multi-dimensional feature-weighted crew-job matching assessment method provided in this application effectively eliminates the coupling interference between shipowner credit levels and crew evaluation scores by constructing a two-way symmetrical credit evaluation framework for both crew members and shipowners and performing covariance decoupling correction on the credit scores of both ends. This ensures that the credit score truly reflects the independent performance responsibility of the evaluation subject, thereby achieving equivalence and fairness in credit evaluation between crew members and shipowners. Simultaneously, this method integrates the decoupled crew credit score as an independent dimension into the multi-dimensional feature vector and adaptively assigns weights to each feature dimension based on job type characteristics and the shipowner's historical operating profile. This allows for dynamic adjustment of the matching focus according to the actual needs of different positions, significantly improving the accuracy of person-job matching. Furthermore, this method pre-verifies the credit threshold for job requirements by decoupling the shipowner's credit score, eliminating invalid positions with unqualified credit from the outset. It also generates a unified matching metric based on dimensional deviation calculation and weighted summation, enabling seamless data flow from credit decoupling to job matching, providing objective and quantifiable technical support for crew recruitment decisions.

[0041] In one embodiment, S1 of the method for evaluating crew member-job matching based on multi-dimensional feature weighting provided by the present invention specifically includes the following steps:

[0042] S11: Obtain multi-dimensional feature data of crew members and ship owners, extract crew members' certificate level, service vessel type, service tonnage, service area and ship owner's enterprise registration information, number of vessels, vessel type and vessel safety certificate level from the multi-dimensional feature data, perform numerical mapping on crew members' certificate level, service vessel type, service tonnage, service area and ship owner's enterprise registration information, number of vessels, vessel type and vessel safety certificate level respectively, convert each qualitative indicator into corresponding quantitative score, and generate crew member static feature coding sequence and ship owner static feature coding sequence;

[0043] S12: Perform time-series normalization on the platform login frequency, resume update cycle and training completion quantity of crew members in the multidimensional feature data, map each time-series indicator to a unified numerical range, generate crew dynamic feature coding sequence, and perform time-series normalization on the recruitment frequency, wage payment on time rate and historical contract completion rate of ship owners in the multidimensional feature data, generate ship owner dynamic feature coding sequence.

[0044] S13: Concatenate the crew static feature coding sequence and the crew dynamic feature coding sequence according to the preset feature dimension order to generate the crew initial feature vector; and concatenate the shipowner static feature coding sequence and the shipowner dynamic feature coding sequence according to the preset feature dimension order to generate the shipowner initial feature vector.

[0045] Specifically, the system acquires multi-dimensional feature data input from crew terminals and shipowner terminals. From the crew's multi-dimensional feature data, it extracts the crew's certificate level, service vessel type, service tonnage, and service area as the raw static feature data. From the shipowner's multi-dimensional feature data, the system extracts the shipowner's company registration information, number of vessels, vessel type, and vessel safety certificate level as the raw static feature data. The system establishes a level sequence mapping relationship for crew certificate levels according to the corresponding competency levels, mapping different levels of certificates to a continuously increasing numerical axis. This mapping relationship is determined based on the certificate's scope of application and legal validity level; the wider the scope of application and the higher the legal validity level, the larger the mapped value. The system establishes a type coding table for service vessel types according to the vessel's operating area attributes and deadweight tonnage, mapping different vessel types to corresponding type code values. The system processes service tonnage using a logarithmic transformation based on actual values. The base of this logarithmic transformation is determined based on the tonnage distribution range to reduce the numerical gap between large and small tonnage.

[0046] The system establishes an ascending mapping relationship for service areas, from coastal areas to unlimited areas, with higher areas receiving larger mapping values. The system extracts two attributes from shipowner registration information: registered capital and years of operation. These attributes are then converted and weighted to form a single enterprise strength score. The weighting coefficients for this score are determined based on the contribution of registered capital and years of operation to the default rate in industry statistics. The system directly assigns numerical values ​​to the number of vessels, encodes vessel types according to a type coding table, and establishes a comprehensive scoring function for vessel safety certificate levels based on certificate validity and coverage. This comprehensive scoring function is the product of the certificate level base score and a monotonically increasing function of the remaining certificate validity period. The system arranges the quantitative scores obtained from the above qualitative indicators in a fixed order to generate static characteristic coding sequences for crew members and shipowners. The formula for calculating the crew certificate level mapping value is as follows:

[0047]

[0048] in, This represents the comprehensive mapping value of the seafarer's certificate. This indicates the total number of certificates held by the crew member. Indicates the first The legal validity weighting coefficient of each certificate is determined based on the level of the issuing authority. Indicates the first The level value of each certificate. Indicates the baseline certificate level value. This represents the attenuation coefficient for grade differences. Preferably, the formula for calculating the shipowner's enterprise strength score is:

[0049]

[0050] in, This indicates the overall strength score of the shipowner company. This indicates the amount of registered capital of the company. This represents the contribution coefficient of the registered capital. Indicates the number of years the company has been in existence. This represents the contribution coefficient of the lifespan, and the fraction is subjected to saturated compression of the lifespan.

[0051] Furthermore, the system extracts the platform login frequency, resume update cycle, and number of training completed by crew members within a preset time window from the crew members' multidimensional feature data as raw data for their dynamic characteristics. Platform login frequency is measured by the number of times a crew member logs in per unit of time; the resume update cycle is measured by the number of days between the crew member's most recent resume update and the current time; and the number of training completed is measured by the number of online training courses completed by the crew member within the preset time window. The system uses a base-2 logarithmic function to compress the crew member's platform login frequency to eliminate the excessive influence of extremely high-frequency login behavior on the numerical distribution. The independent variable of this logarithmic function is the number of logins plus one to ensure that there is a defined value for zero logins. The system uses a negative exponential decay function to transform the crew member's resume update cycle. The independent variable of this decay function is the update interval in days, and the decay coefficient is determined based on the crew member's average contract period. The shorter the update interval, the larger the transformed value, reflecting the crew member's proactive behavior of keeping their resume information updated in a timely manner. The system uses a saturated linear function to process the number of training sessions completed. This function grows linearly when the number of training sessions is below a preset threshold, and the growth rate decreases until saturation after the number of training sessions exceeds the preset threshold, in order to prevent overtraining from taking an unreasonable weight in the dynamic features.

[0052] The system normalizes the values ​​of the three time-series indicators after processing by their respective transformation functions. Normalization uses the range standardization method, taking the minimum and maximum observed values ​​of each indicator as the standardization boundary, mapping the indicator values ​​to a unified numerical range, and generating a crew dynamic feature coding sequence. The system extracts the shipowner's recruitment frequency, wage payment on-time rate, and historical contract completion rate from the shipowner's multi-dimensional feature data as the original data for shipowner dynamic features. Recruitment frequency is measured by the number of job postings published by the shipowner per unit time; wage payment on-time rate is measured by the proportion of on-time wage payments recorded through the platform to the total number of payments; and historical contract completion rate is measured by the proportion of completed contracts to the total number of signed contracts. The system applies logarithmic compression to the recruitment frequency, the same as the crew login frequency, and normalizes the wage payment on-time rate and historical contract completion rate using the original proportional values, generating a shipowner dynamic feature coding sequence. The conversion formula for login frequency in crew dynamic features is:

[0053]

[0054] in, This represents the login frequency conversion value. This represents the number of logins per unit of time. The formula for converting the resume update cycle is:

[0055]

[0056] in, This represents the timeliness of resume updates, with a value ranging from zero to one. This represents the time decay factor, with its dimension being the reciprocal of the unit of time. It is set based on the average contract length of crew members. This represents the number of days since the last update, measured in units of time. The formula for handling saturation in training completion counts is:

[0057]

[0058] in, This represents the training saturation value. Indicates the number of training sessions completed. This represents the saturation threshold constant.

[0059] Further, the system acquires the static feature encoding sequence and the dynamic feature encoding sequence of the crew members, and concatenates the two sequences end-to-end according to a preset feature dimension order. The preset feature dimension order places all static feature encoding components at the beginning and all dynamic feature encoding components at the end, with each component occupying a fixed position in the concatenated sequence. The resulting initial feature vector of the crew members is a one-dimensional real-number vector. The total dimension of this vector is equal to the sum of the dimensions of the static and dynamic feature encoding sequences of the crew members. The order of the components in the vector is completely consistent with the preset dimension order, and the value of each component is the quantized value of the corresponding feature after processing in the aforementioned steps. During the concatenation process, the system records the index position of each component in the vector and the corresponding feature name, establishing an index mapping table to support the subsequent steps of locating and adjusting the weights of specific feature components. The system acquires the shipowner's static feature encoding sequence and dynamic feature encoding sequence. Following the same concatenation logic as the crew feature vector, it concatenates the static feature encoding sequence first and the dynamic feature encoding sequence last to generate the initial shipowner feature vector. The system then associates and stores the generated initial crew and shipowner feature vectors with the crew identification identifier and shipowner identification identifier, respectively. The concatenation formula for the initial crew feature vector is:

[0060]

[0061] in, This represents the initial feature vector of the crew. This represents the static feature encoding sequence vector of the crew. Represents the dynamic feature encoding sequence vector of crew members, with superscript... This represents the transpose of a vector, where two vectors are concatenated vertically. The formula for concatenating the initial feature vectors of the shipowner is:

[0062]

[0063] in, This represents the shipowner's initial feature vector. This represents the static feature encoding sequence vector of the shipowner. This represents the shipowner's dynamic feature encoding sequence vector. The above concatenation operation requires... and Having the same data type and numerical precision, and They have the same data type and numerical precision to ensure consistency of vectors during numerical operations.

[0064] In one embodiment, S2 of the method for evaluating crew member-job matching based on multi-dimensional feature weighting provided by the present invention specifically includes the following steps:

[0065] S21: Extract the shipowner's evaluation score, anonymous peer score of fellow crew members, and platform objective behavior evaluation score after the end of each contract from the historical evaluation data of crew members in the multidimensional feature data. Weight the shipowner's evaluation score, anonymous peer score of fellow crew members, and platform objective behavior evaluation score according to the preset first weight coefficient, second weight coefficient, and third weight coefficient to generate the crew's comprehensive score for each contract.

[0066] The sum of the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient is 1.

[0067] S22: Based on the time interval between the crew member's comprehensive score for each contract and the current time, determine the time decay weight coefficient corresponding to each contract. Then, calculate the weighted sum of the crew member's comprehensive score for each contract and the time decay weight coefficient, and divide the sum by the time decay weight coefficient to generate the crew member's initial credit score.

[0068] S23: Extract the crew's evaluation scores of the shipowner from the shipowner's historical evaluation data in the multidimensional feature data after the end of each contract. Take the average of multiple crew evaluation scores within the same contract period as the shipowner's single contract comprehensive score for that contract. Calculate the shipowner's single contract comprehensive score using a weighted moving average method that is the same as the crew's initial credit score, and generate the shipowner's initial credit score.

[0069] Specifically, the system extracts historical evaluation records of crew members after the end of each contract from multi-dimensional feature data. For each contract cycle, the system locates the shipowner's evaluation record corresponding to the contract. This record contains the shipowner's scores for the crew member in five dimensions: professional skills, work responsibility, obedience to management, work efficiency, and equipment maintenance. The system sums these scores in an equally weighted manner as the initial value for the shipowner's evaluation score. The system also locates the anonymous peer evaluation record of the crew members on the same ship corresponding to the contract. This record contains scores from other crew members on the same ship for the crew member in five dimensions: teamwork, communication skills, hygiene habits, interpersonal relationships, and helpfulness. After removing the highest and lowest scores in these five dimensions, the system takes the arithmetic mean of the scores in the remaining three dimensions as the anonymous peer evaluation score. Finally, the system locates the platform's objective behavior evaluation record corresponding to the contract. This record contains three indicators: contract fulfillment rate, on-time reporting rate, and community contribution value.

[0070] The system calculates the contract performance rate as the ratio of actual days on board to the number of days stipulated in the contract; the on-time reporting rate as the ratio of the number of on-time reports to the total number of reports required; and the community contribution value as the number of valid experience-sharing posts published by crew members in the platform community. The system standardizes these three indicators and sums them using equal weighting to obtain the platform's objective behavior evaluation score. The system multiplies the shipowner's evaluation score by a first weighting coefficient, the anonymous peer review score by a second weighting coefficient, and the platform's objective behavior evaluation score by a third weighting coefficient. These three products are then summed to generate the crew member's comprehensive score for each contract. The values ​​of the first, second, and third weighting coefficients are all between zero and one, and their sum is one. The formula for calculating the crew member's comprehensive score for a single contract is as follows:

[0071]

[0072] in, Indicates the crew members at the The overall score for a single contract after the conclusion of this contract is a dimensionless score with the same dimension as each individual evaluation score. , , Let these represent the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient, respectively. All three are dimensionless real numbers and satisfy the following condition: The constraints, Indicates the first The shipowner's evaluation score in this contract, calculated using a weighted summation of five dimensions. Indicates the first The anonymous peer ratings of the same crew members in this contract, after removing extreme values ​​and averaging, are as follows. Indicates the first The platform's objective behavior evaluation score for each contract is calculated by standardizing and summing three indicators. The system associates and stores the crew member's comprehensive score for each contract with the contract's end timestamp, forming a historical comprehensive score sequence for that crew member.

[0073] Further, the system obtains the sequence of comprehensive scores for each individual contract corresponding to all historical contracts of the crew member, as well as the contract end timestamps for each contract. Using the current system time as the base time, the system calculates the time interval (in days) between the end timestamp of each contract and the current time. For each time interval, the system calculates the time decay weight coefficient for the corresponding contract using a time decay function. This time decay function is a monotonically decreasing function; the larger the time interval, the smaller the time decay weight coefficient, and vice versa. The system multiplies the comprehensive score of each individual contract by the corresponding time decay weight coefficient to obtain the weighted score for each contract. The system sums the weighted scores of all contracts to obtain the total weighted score. The system sums the time decay weight coefficients of all contracts to obtain the total time decay weight coefficient. The system divides the total weighted score by the total time decay weight coefficient to generate the crew member's initial credit score. The formula for calculating the time decay weight coefficient is:

[0074]

[0075] in, Indicates the first The time decay weighting coefficient for this contract is a dimensionless real number, ranging from zero to one. This represents the decay rate constant, with dimensions in the negative square of a unit of time, used to control the rate of decay over time. Indicates the first The time interval between the contract end timestamp and the current time, expressed in units of time. The physical meaning of this formula is that as the time interval increases squared, the weighting coefficient decreases inversely, ensuring that the most recently occurring contract has the highest contribution to the current credit score, while the contribution of earlier contracts gradually decreases. The formula for calculating a seafarer's initial credit score is:

[0076]

[0077] in, This represents the crew member's initial credit score, with the same dimension as the overall score for a single contract; it is a dimensionless score. This represents the total number of historical contracts for the crew member, a dimensionless positive integer. The numerator represents the sum of the individual contract scores weighted by the time decay weighting factor, and the denominator represents the sum of all time decay weighting factors. This division operation ensures that the weighted average credit score remains within the original range of the individual contract scores. The system writes the calculated initial credit score of the crew member into the crew member credit table.

[0078] Furthermore, the system extracts historical evaluation records of shipowners after the end of each contract from multidimensional feature data. For each contract period of a shipowner, the system locates all crew evaluation records of the shipowner corresponding to that contract. Each evaluation record includes scores from the crew members involved in the contract regarding nine dimensions of the shipowner: on-time and full wage payment rate, contract terms fulfillment, welfare benefits delivery, ship safety and maintenance, work intensity and leave arrangements, living conditions and food standards, professionalism of the recruitment process, management communication and humanistic care, and fair treatment and discriminatory behavior. The system takes the arithmetic mean of the scores for the same dimension in all crew evaluation records within the contract period to obtain the average score for each dimension within that contract period. Then, the average scores of the nine dimensions are summed in an equally weighted manner to obtain the shipowner's comprehensive score for that contract. The system obtains the sequence of comprehensive scores for all historical contracts of the shipowner and the contract end timestamps for each contract. The system uses the current system time as the base time and adopts the same time decay weighting calculation method as the initial credit score of the crew members, i.e., according to... Calculate the time decay weighting coefficient for each contract, where This indicates that the shipowner's number The time interval between the end timestamp of this contract and the current time. The same constant decay rate value used in the initial credit score calculation for crew members is adopted. The system multiplies the shipowner's overall score for each contract by the corresponding time decay weighting coefficient, then sums the results to obtain the shipowner's weighted score. This weighted score is then divided by the sum of all time decay weighting coefficients to generate the shipowner's initial credit score. The formula for calculating the shipowner's initial credit score is as follows:

[0079]

[0080] in, This represents the shipowner's initial credit score, measured as a dimensionless score. This represents the total number of historical contracts held by the shipowner, and is a dimensionless positive integer. This indicates that the shipowner's number The overall score for each contract is a dimensionless score, calculated by weighted summation of the nine-dimensional scores given by all crew members to the shipowner in that contract. This indicates the time decay weighting coefficient corresponding to this contract. The system will then write the calculated initial credit score of the shipowner into the shipowner credit table.

[0081] In one embodiment, S3 of the method for evaluating crew member-job matching based on multi-dimensional feature weighting provided by the present invention specifically includes the following steps:

[0082] S31: Using each employment-evaluation record in the historical evaluation data of crew members and shipowners in the multidimensional feature data as edges, and each crew member and each shipowner as nodes, construct a bipartite graph of evaluation relationships;

[0083] Each edge connects a crew member node to a ship owner node and carries a first weight value and a second weight value. The first weight value is the ship owner's evaluation score of the crew member, and the second weight value is the crew member's evaluation score of the ship owner.

[0084] S32: For each edge in the bipartite graph, calculate the average of the first weight values ​​of the shipowner nodes connected to the corresponding edge with respect to all historically connected crew members as the first benchmark value, calculate the average of the second weight values ​​of the crew member nodes connected to the corresponding edge with respect to all historically connected shipowners as the second benchmark value, and multiply the absolute value of the difference between the first weight value and the first benchmark value by the absolute value of the difference between the second weight value and the second benchmark value to generate the local coupling coefficient of the corresponding edge.

[0085] S33: For each crew member node, sum the local coupling coefficients of all edges connected to the crew member node and divide the sum by the total number of edges connected to the crew member node to generate the global coupling strength of the crew member node; for each ship owner node, sum the local coupling coefficients of all edges connected to the ship owner node and divide the sum by the total number of edges connected to the ship owner node to generate the global coupling strength of the ship owner node.

[0086] Specifically, the system extracts evaluation records generated within each employment cycle from historical evaluation data of crew members and shipowners as the basis for graph construction. Each employment cycle corresponds to a labor contract between the crew member and the shipowner. After the contract is completed, the system records the shipowner's evaluation scores for the crew member across multiple dimensions, and the crew member's evaluation scores for the shipowner across multiple dimensions. The system treats each crew member as a first-type node and each shipowner as a second-type node, forming two disjoint sets of nodes in a bipartite graph. For each employment evaluation record, the system establishes an undirected edge between the corresponding crew member node and shipowner node, logically carrying evaluation information in both directions simultaneously. The system uses the shipowner's evaluation score for the crew member as the first weight value of this edge, and the crew member's evaluation score for the shipowner as the second weight value. Both the first and second weight values ​​are real numbers and stored independently. When there are multiple employment records between the same crew member node and the same shipowner node, the system creates an independent edge for each employment record. This means multiple parallel edges are allowed between the same node pair. Each edge carries a first weight value and a second weight value corresponding to its respective contract period. After traversing all historical evaluation records, the system outputs a bipartite graph data structure containing all crew member nodes, all shipowner nodes, and all edges. The first weight value sequence and the second weight value sequence of each edge in this graph are recorded in the edge's attribute field.

[0087] Furthermore, the system traverses each edge in the bipartite graph of evaluation relationships, calculating benchmark values ​​for the shipowner and crew nodes connected to that edge. The system obtains all historical connections to the shipowner node connected to the current edge (i.e., all edges connected to the shipowner node as one end), extracts the first weight values ​​of these edges to form a first weight value set corresponding to the shipowner node, and calculates the arithmetic mean of this set as the first benchmark value. This first benchmark value reflects the shipowner's average tendency to give evaluation scores to hired crew members in previous employments. The system also obtains all historical connections to the crew node connected to the current edge (i.e., all edges connected to the crew node as one end), extracts the second weight values ​​of these edges to form a second weight value set corresponding to the crew member node, and calculates the arithmetic mean of this set as the second benchmark value. This second benchmark value reflects the crew member's average tendency to give evaluation scores to the shipowners they serve in previous employments. Finally, the system calculates the difference between the first weight value and the first benchmark value of the current edge and takes the absolute value to obtain the shipowner evaluation deviation for that edge. The system calculates the absolute value of the difference between the second weight value and the second benchmark value of the current edge to obtain the crew evaluation deviation for that edge. The system then multiplies the shipowner evaluation deviation by the crew evaluation deviation to obtain the local coupling coefficient for that edge. The formula for calculating the local coupling coefficient is:

[0088]

[0089] in, Indicates the first The local coupling coefficient of the strip, with dimensions equal to the square of the evaluation score. Indicates the first The shipowner's evaluation score for the crew on the side of the label is the first weighted value, which is a dimensionless score. This represents the average of the first weight values ​​of all historical edges connected to the shipowner node by this edge, i.e., the first baseline value, with the same dimensions as the first weight value. Indicates the first The second weighted score, which represents the crew's evaluation of the shipowner, is a dimensionless score. This represents the average of the second weight values ​​(i.e., the second baseline value) of the crew member nodes connected by this edge across all historical edges, with the same dimensions as the second weight value. The physical meaning of this formula is that when a shipowner's evaluation of a crew member significantly deviates from the shipowner's average evaluation style, and simultaneously, the crew member's evaluation of the shipowner significantly deviates from the crew member's average evaluation style, there is a high probability of bidirectional abnormal coupling in the employment relationship corresponding to this edge.

[0090] Further, after calculating the local coupling coefficients of all edges, the system performs a global coupling strength aggregation operation on each node in the bipartite graph of the evaluation relationship. For each crew member node, the system locates all edges connected to the crew member node, extracts the local coupling coefficients of these edges, and sums all the extracted local coupling coefficients to obtain the total local coupling coefficient of the crew member node. The system counts the total number of connected edges of the crew member node, which is equal to the number of employment contracts the crew member has participated in throughout history. The system divides the total local coupling coefficient by the total number of connected edges to obtain the global coupling strength of the crew member node, which represents the average degree of bidirectional evaluation deviation between the crew member and different shipowners in all historical employment relationships. For each shipowner node, the system locates all edges connected to the shipowner node, extracts the local coupling coefficients of these edges, and sums all the extracted local coupling coefficients and divides them by the total number of connected edges of the shipowner node to obtain the global coupling strength of the shipowner node. The calculation formulas for the global coupling strength of the crew member node and the global coupling strength of the shipowner node are as follows:

[0091]

[0092]

[0093] in, Indicates the first The global coupling strength of each crew member node, whose dimensions are consistent with those of the local coupling coefficient, is the square of the evaluation score dimension. This represents the total number of connected edges of the crew member node, a dimensionless positive integer. This represents the set of all edges connected to the crew member node. Indicates the first The global coupling strength of each shipowner node has the same dimensions as the local coupling system. This represents the total number of connected edges of the shipowner node, a dimensionless positive integer. This represents the set of all edges connected to the shipowner node. The physical meaning of this aggregation operation is that, through the cumulative averaging of local coupling coefficients at the edge level, the degree of bidirectional evaluation anomaly in a single employment relationship is extended to the node level, reflecting the coupling strength level of the crew member or shipowner in the overall evaluation network, providing a quantitative basis at the node dimension for subsequent significant coupling determination. The system writes the calculated global coupling strength values ​​of each crew member node and each shipowner node into the node attribute field, respectively.

[0094] In one embodiment, S4 of the method for evaluating crew member-job matching based on multi-dimensional feature weighting provided by the present invention specifically includes the following steps:

[0095] S41: Based on the comparison results of global coupling strength and preset coupling threshold, the initial credit scores of all shipowner nodes connected to the crew node that is judged to have significant coupling are extracted as the initial credit score sequence of shipowners. The first weight value obtained by the crew node on each connection edge is extracted as the shipowner evaluation score sequence corresponding to the crew. The ratio of the covariance between the initial credit score sequence of shipowners and the evaluation score sequence of shipowners to the variance of the initial credit score sequence of shipowners is calculated to generate the shipowner factor interference coefficient.

[0096] S42: Calculate the average of the initial credit scores of all shipowner nodes connected to the crew node as the crew's historical credit average. Obtain the global average of the initial credit scores of all shipowner nodes on the platform as the global credit average. Multiply the shipowner factor interference coefficient by the difference between the historical credit average and the global credit average. Subtract the product of the shipowner factor interference coefficient and the difference between the historical credit average and the global credit average from the crew's initial credit score of the crew node to generate the crew decoupling credit score of the crew node.

[0097] S43: Based on the crew factor interference coefficient, for the shipowner node that is determined to have significant coupling, calculate the crew factor interference coefficient according to the initial credit score sequence of all crew nodes connected to the shipowner node and the crew evaluation score sequence corresponding to the shipowner. Use the crew factor interference coefficient to correct the initial credit score of the shipowner node and generate the shipowner decoupling credit score of the shipowner node.

[0098] Specifically, the system filters all crew member nodes based on a comparison between the global coupling strength and a preset coupling threshold, extracting those with a global coupling strength exceeding the preset threshold as crew member nodes exhibiting significant coupling. For each crew member node identified as having significant coupling, the system locates all shipowner nodes connected to that node, extracts the initial credit scores of these shipowner nodes, and forms a sequence of initial credit scores for that crew member node. The system locates the first weight value of that crew member node on each connection edge, i.e., the shipowner's evaluation score for the crew member, forming a sequence of shipowner evaluation scores for that crew member node. The order of the components in this sequence corresponds one-to-one with the order of the shipowner nodes in the initial credit score sequence. The system calculates the arithmetic mean of the initial credit score sequence as the mean of the initial credit score sequence, and calculates the arithmetic mean of the shipowner evaluation score sequence as the mean of the shipowner evaluation score sequence.

[0099] Preferably, the system calculates the deviation of each component in the shipowner's initial credit score sequence from the sequence mean, and the deviation of each component in the shipowner's evaluation score sequence from the sequence mean. The deviations at corresponding positions in the two sequences are multiplied and summed to obtain the covariance of the two sequences. The system calculates the sum of squares of the deviations of each component in the shipowner's initial credit score sequence from the sequence mean to obtain the variance of the shipowner's initial credit score sequence. The system divides the covariance by the variance to obtain the shipowner factor interference coefficient. The formula for calculating the shipowner factor interference coefficient is:

[0100]

[0101] in, This represents the interference coefficient caused by shipowner factors. Its dimension is the ratio of the evaluation score to the credit score. It is used to indicate the average change in the crew evaluation score corresponding to a one-unit change in the shipowner's credit score. This represents the total number of shipowner nodes connected to the crew member node, and is a dimensionless positive integer. Indicates the first The initial credit score of each shipowner node, measured in units of credit score. This represents the mean of the sequence, with units of and . Consistent Indicates the first The shipowner's evaluation score for the crew on the connecting side is the first weighted value, and the dimension is the unit of evaluation score. This represents the mean of the sequence, with units of and . Consistent. The physical meaning of this formula lies in quantifying the linear driving force of shipowner creditworthiness on the evaluation score through the least-squares linear regression slope.

[0102] Furthermore, the system obtains the initial credit scores of all shipowner nodes connected to the crew member node, calculates the arithmetic mean of these initial credit scores, and uses it as the historical credit mean of the shipowners for that crew member node. The system also obtains the initial credit scores of all shipowner nodes on the platform, calculates their arithmetic mean, and uses it as the global credit mean of the shipowners. The system calculates the difference between the historical credit mean and the global credit mean of the shipowners, reflecting the degree of deviation of the credit level of the shipowner group served by the crew member historically relative to the average credit level of shipowners across the platform. The system multiplies the shipowner factor interference coefficient by this difference to obtain a credit correction amount. The system subtracts this credit correction amount from the initial credit score of the crew member node to generate the decoupled credit score of the crew member node. The logic of this correction operation is that when the credit level of the shipowners served by the crew member historically is generally higher than the platform average, the crew member's evaluation score may be overestimated due to the positive interference of the high credit of the shipowners, therefore, the corresponding interference amount needs to be deducted from the initial credit score; conversely, when the credit level of the shipowners is generally lower than the platform average, the corresponding interference amount needs to be compensated. The formula for calculating the crew decoupling credit score is as follows:

[0103]

[0104] in, This represents the decoupled credit score for crew members, with the same dimensions as the initial credit score, and serves as the unit of credit score. This represents the initial credit score of the crew member, measured in units of credit score. This represents the interference coefficient caused by shipowner factors, with the dimension being the ratio of the evaluation score to the credit score. This represents the average of the initial credit scores of all shipowner nodes connected to the crew member, i.e., the historical average credit score of the shipowners, measured in credit score units. This represents the average initial credit score of all shipowner nodes across the entire platform, i.e., the global credit average of shipowners, measured in credit score units. In this formula... The product of the difference in parentheses and the product of the two values ​​is expressed in units of credit score. The dimensions are consistent, ensuring the physical validity of the subtraction operation.

[0105] Furthermore, the system performs a symmetric decoupling correction operation on shipowner nodes identified as having significant coupling. The system locates all crew member nodes connected to the shipowner node, extracts the initial credit scores of these crew members, and forms a sequence of initial credit scores for the crew members corresponding to that shipowner node. The system locates the second weight value of the shipowner node on each connection edge, i.e., the crew member's evaluation score of the shipowner, and forms a sequence of crew member evaluation scores for that shipowner node. The system calculates the mean of the initial credit score sequence and the mean of the crew member evaluation score sequence, and calculates the ratio of the covariance of the two sequences to the variance of the initial credit score sequence to obtain the crew member factor interference coefficient. This coefficient quantifies the linear influence of the crew member's credit level on the evaluation score obtained by the shipowner. The system calculates the average of the initial credit scores of all crew member nodes connected to the shipowner node as the historical average credit score of the crew member for that shipowner node. The system obtains the global average of the initial credit scores of all crew member nodes on the platform as the global average credit score of the crew members. The system calculates the difference between the historical average credit score of crew members and the global average credit score of crew members, and then multiplies this difference by the crew member factor interference coefficient to obtain the shipowner credit correction amount. The system subtracts this correction amount from the shipowner's initial credit score for that shipowner node to generate the shipowner decoupled credit score for that shipowner node. The formula for calculating the shipowner decoupled credit score is as follows:

[0106]

[0107] in, This represents the shipowner's decoupled credit score, with the same dimensions as the shipowner's initial credit score, and is a unit of credit score. This represents the shipowner's initial credit score, measured in units of credit points. The crew factor interference coefficient is represented by the following formula: ,in This indicates the total number of crew nodes connected to the shipowner node. Indicates the first The initial credit score of each crew member node. This represents the mean of the sequence. Indicates the first The second weighted value is the score given by the crew members on the connecting side to the shipowner. This represents the mean of the sequence. The unit of measurement is the ratio of the evaluation score to the credit score. This represents the average initial credit score of all crew members connected to the shipowner, i.e., the average historical credit score of the crew members, measured in credit score units. This represents the average initial credit score of all crew members across the entire platform, i.e., the global average credit score of all crew members, measured in units of credit score. This symmetric correction system ensures that the credit scores of both crew members and shipowners are independent of the coupling interference from each other's credit levels.

[0108] In one embodiment, S5 of the method for evaluating crew member-job matching based on multi-dimensional feature weighting provided by the present invention specifically includes the following steps:

[0109] S51: Obtain the job requirement data of the current job to be matched from the shipowner terminal, extract the minimum credit score requirement field from the job requirement data, compare the minimum credit score requirement with the shipowner's decoupled credit score, if the shipowner's decoupled credit score is lower than the minimum credit score requirement, mark the job requirement as not meeting the credit threshold and remove the job requirement from the matching candidate set; if the shipowner's decoupled credit score is greater than or equal to the minimum credit score requirement, retain the job requirement.

[0110] S52: The certificate level requirements, service experience requirements, and special skill requirements in the retained job requirement data are vectorized and encoded according to the same encoding rules as the initial feature vector of the crew members, to generate a job requirement feature vector;

[0111] S53: Perform absolute value calculation on the difference between the values ​​of each dimension in the crew member multi-dimensional feature weighted vector and the values ​​of the corresponding dimensions in the job requirement feature vector, and subtract the absolute value of the difference of each dimension from 1 to generate the basic matching degree of each dimension.

[0112] S54: Based on the job type characteristics and ship type characteristics of the current position, match the corresponding basic weight vector from the preset weight configuration scheme, and adjust the bias of the basic weight vector according to the shipowner's historical operating characteristics in the shipowner's initial feature vector to generate adaptive weights for each dimension. Multiply the basic matching degree of each dimension with the adaptive weight of the corresponding dimension and sum them to generate a weighted comprehensive matching degree.

[0113] Specifically, the system receives job requirement data for the currently matched positions from the shipowner's terminal. This data is stored in the matching request message sent by the shipowner's terminal in the form of structured fields. The system parses the minimum credit score requirement field from the job requirement data. The value of this field is a floating-point number, with the same dimensions as the credit score. The system reads the corresponding shipowner decoupled credit score from the shipowner's credit table. This score is a value after covariance decoupling correction, with the same dimensions as the minimum credit score requirement. The system performs a numerical comparison operation, determining the value of the shipowner's decoupled credit score against the value of the minimum credit score requirement. When the shipowner's decoupled credit score is less than the minimum credit score requirement, the system determines that the job requirement does not meet the credit threshold condition, removes the matching record corresponding to the job requirement from the matching candidate set, and the job requirement does not participate in any subsequent matching degree calculation process. No feature data of the job requirement is retained in the matching candidate set. When the shipowner's decoupled credit score is greater than or equal to the minimum credit score requirement, the system determines that the job requirement meets the credit threshold condition, and retains all data records of the job requirement in the matching candidate set. The system performs the above comparison and filtering operations on each job requirement in the candidate set until only job requirements meeting the credit threshold are retained. The mathematical expression for the credit threshold determination criteria is:

[0114]

[0115] in, This indicates whether the credit threshold has been met; a value of one indicates compliance, and a value of zero indicates non-compliance. This is an indicator function; it takes the value of one when the condition within the parentheses is true, and the value of zero when the condition is false. This represents the shipowner's decoupled credit score, measured in units of credit score. This indicates the minimum credit score requirement for the job posting, with units of... Consistent. The result of this judgment is the basis for selecting the matching candidate set.

[0116] Furthermore, the system performs vectorized encoding on all remaining job requirements in the candidate matching set. The system extracts the certificate level requirement field, service experience requirement field, and special skills requirement field from each job requirement data as key attributes of the job requirement. The certificate level requirement field includes the minimum required certificate level and the range of vessel types covered by the certificate. The service experience requirement field includes the minimum required years of service, minimum service area level, and minimum service vessel tonnage level. The special skills requirement field includes additional skills required for the job, including but not limited to specific vessel operation qualifications and specific equipment operation qualifications. The system uses the exact same encoding rules as the initial feature vector of the crew members to convert the above attributes into numerical components. For certificate level requirements, the system converts the certificate level requirement into the corresponding level value according to the same level value table mapping to the crew member's certificate level; for service experience requirements, the system converts the years of service, service area level, and tonnage level into corresponding quantified values ​​according to the same mapping function as the crew member's service experience; for special skills requirements, the system encodes the existence of each skill into binary components according to a preset skills encoding table. The system arranges the above components in the same feature dimension order as the initial feature vector of the crew members, forming the job requirement feature vector. The first part of the job requirement feature vector... The encoding formula for the dimensional component is:

[0117]

[0118] in, The first element representing the feature vector of job requirements The dimension component has the same dimensions as the corresponding dimension in the initial feature vector of the crew member, and is a dimensionless numerical value. This indicates the first [item] in the job requirement data. The original requirement attribute values ​​of each dimension, Indicates the first The corresponding encoding mapping function for dimension , which is related to the first dimension in the initial feature vector of the crew. The encoding mapping functions for each dimension are completely identical, ensuring that the values ​​of the two vectors are comparable in the same dimension. The system associates and stores the encoded job requirement feature vectors with the job identifiers.

[0119] Furthermore, the system obtains the job requirement feature vectors corresponding to each job requirement in the matching candidate set, as well as the crew member multi-dimensional feature weighted vectors corresponding to each crew member. The crew member multi-dimensional feature weighted vector is formed by concatenating the initial crew member feature vector and the crew member decoupled credit score. Its total number of dimensions is equal to the number of dimensions of the initial crew member feature vector plus one, where the last dimension is the crew member decoupled credit score. The number of dimensions of the job requirement feature vector is the same as the number of dimensions of the initial crew member feature vector, excluding the credit score dimension. When performing matching degree calculation, the system aligns the job requirement feature vector with each dimension of the crew member multi-dimensional feature weighted vector except for the last dimension. For each dimension, the system calculates the absolute value of the difference between the component value of the job requirement feature vector in that dimension and the component value of the crew member multi-dimensional feature weighted vector in that dimension. This absolute value represents the degree of absolute deviation between the crew member's actual performance in that dimension and the job requirements. The system subtracts this absolute value from the numerical value to obtain the basic matching degree for that dimension. The basic matching degree takes a value between zero and one, with a value of one when the crew member's performance is completely consistent with the job requirements and a value of zero when the deviation reaches its maximum. The formula for calculating the basic matching degree is:

[0120]

[0121] in, Indicates the first The basic matching degree of dimension is a dimensionless real number, ranging from zero to one. The weighted vector of multi-dimensional features of crew members is represented in the th... Dimensional component values, The feature vector representing the job requirements is in the th position. The component values ​​of the dimension are both dimensionless, and the absolute value of the difference has the same dimension as both. Therefore... It is a dimensionless quantity. This formula converts the absolute deviation into a relative degree of fit through complementary normalization.

[0122] Furthermore, the system matches the corresponding basic weight vector from a preset weight configuration scheme based on the job type characteristics and ship type characteristics of the current position. The preset weight configuration scheme is stored in the system's weight configuration table, which uses the job type identifier and ship type identifier as a joint primary key. Each primary key corresponds to a basic weight vector with the same number of dimensions as the initial feature vector of the crew member. The system extracts the job type identifier and ship type identifier of the current position and performs precise matching in the weight configuration table to obtain the basic weight vector corresponding to that position. The system extracts the shipowner's historical operating characteristic components from the shipowner's initial feature vector. These components include the shipowner's historical recruitment frequency, historical contract completion rate, and historical average employment duration on the platform. The system adjusts the bias of each dimension in the basic weight vector according to the shipowner's historical operating characteristics. The adjustment method is to multiply the basic weight value of each dimension by a bias factor. The bias factor is calculated by the shipowner's historical operating characteristics using a preset bias function; the bias factor can be different for different dimensions. The system normalizes the adjusted weight values ​​of each dimension so that the sum of the weights of each dimension equals one, generating adaptive weights for each dimension. The system multiplies the base matching score of each dimension by the corresponding adaptive weight to obtain a weighted matching score component for each dimension. The weighted matching score components of all dimensions are then summed to generate a weighted overall matching score. The formula for calculating the weighted overall matching score is:

[0123]

[0124] in, This represents the weighted overall matching degree, a dimensionless real number ranging from zero to one. This represents the total number of feature dimensions, a dimensionless positive integer. Indicates the first The basic matching degree of the dimension, Indicates the first The adaptive weights of all dimensions are dimensionless real numbers. The sum of them equals one. Indicating the weight configuration scheme, the first... The fundamental weight values ​​of the dimension are dimensionless real numbers. Indicates the first The bias factor of dimension is a dimensionless real number and is a vector of historical operating characteristics of the shipowner. The function, It includes three components: shipowner's historical recruitment frequency, contract completion rate, and average employment duration. The dimensions of each component are the reciprocal of time, dimensionless proportion, and time unit, respectively. The output is a dimensionless real number. In this formula, the numerator calculates the bias-adjusted weight value for each dimension, and the denominator sums and normalizes over all dimensions to ensure the normalization constraint of the adaptive weights. The system outputs the weighted comprehensive matching degree to the job matching result list as the sorting basis.

[0125] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0126] Based on the same inventive concept, this application also provides a multi-dimensional feature-weighted crew-job matching evaluation device for implementing the aforementioned multi-dimensional feature-weighted crew-job matching evaluation method. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations of one or more embodiments of the multi-dimensional feature-weighted crew-job matching evaluation device provided below can be found in the limitations of the multi-dimensional feature-weighted crew-job matching evaluation method described above, and will not be repeated here.

[0127] Preferably, such as Figure 2 As shown, this invention provides a crew member and job matching evaluation system 600 based on multi-dimensional feature weighting. This system is configured with the following modules:

[0128] The multidimensional feature encoding module 610 is used to acquire multidimensional feature data of crew members and ship owners, perform vectorization encoding on the multidimensional feature data, and generate initial feature vectors for crew members and ship owners.

[0129] The main credit calculation module 620 is used to calculate the comprehensive score of a single contract for a crew member by weighted summing of the multi-party evaluation scores in the historical evaluation data in the multi-dimensional feature data, and to generate the initial credit score of the crew member and the initial credit score of the shipowner by weighted moving average of the comprehensive score of the single contract for a crew member.

[0130] The bipartite graph coupling strength solution module 630 is used to construct a bipartite graph of the evaluation relationship between crew members and ship owners based on historical evaluation data. It performs deviation product calculation on the evaluation score on each edge of the bipartite graph and the initial credit scores of the crew members and ship owners at the corresponding two ends of the node to generate the global coupling strength between each crew member node and each ship owner node.

[0131] The credit score covariance decoupling module 640 is used to filter out the crew credit scores and shipowner credit scores with significant coupling based on the global coupling strength, and to perform covariance decoupling correction processing on the crew credit scores and shipowner credit scores with significant coupling respectively to generate crew decoupled credit scores and shipowner decoupled credit scores.

[0132] The 650 comprehensive matching calculation module for personnel and positions is used to perform credit threshold verification on job requirement data obtained from the shipowner terminal based on the shipowner's decoupled credit score. It compares the minimum credit score requirement of the job requirement with the shipowner's decoupled credit score, eliminates invalid job requirements that do not meet the credit threshold, and outputs the job requirement feature vector. It concatenates the initial feature vector of the crew member with the crew member's decoupled credit score to generate a multi-dimensional feature weighted vector of the crew member. It combines the job requirement feature vector with the multi-dimensional feature weighted vector of the crew member to perform dimensional deviation calculation and weighted summation to generate a weighted comprehensive matching degree between the crew member and the job.

[0133] The weighted overall matching degree is used to indicate the degree of suitability between the crew member and the current position. The higher the matching degree, the more suitable the crew member is for the current position.

[0134] In one embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method for evaluating crew-job matching based on multi-dimensional feature weighting.

[0135] In one embodiment, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for evaluating crew-job matching based on multi-dimensional feature weighting.

[0136] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0137] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0138] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for evaluating the matching degree between crew members and positions based on multi-dimensional feature weighting, characterized in that, Includes the following steps: S1: Obtain multi-dimensional feature data of crew members and ship owners, and perform vectorization encoding on the multi-dimensional feature data to generate initial feature vectors for crew members and ship owners; S2: Based on the historical evaluation data in the multidimensional feature data, the multi-party evaluation scores in the historical evaluation data are weighted and summed to calculate the comprehensive score of a single contract for the crew member, and the comprehensive score of a single contract for the crew member is weighted and moved averaged to generate the initial credit score of the crew member and the initial credit score of the shipowner. S3: Based on the historical evaluation data, construct a bipartite graph of the evaluation relationship between crew members and ship owners. Perform a deviation product operation on the evaluation score on each edge of the bipartite graph and the initial credit scores of the crew members and the ship owners at the corresponding two ends of the node to generate the global coupling strength between each crew member node and each ship owner node. S4: Based on the global coupling strength, select the crew credit scores and shipowner credit scores that have significant coupling, and perform covariance decoupling correction processing on the crew credit scores and shipowner credit scores that have significant coupling to generate crew decoupled credit scores and shipowner decoupled credit scores. S5: Based on the shipowner's decoupled credit score, perform credit threshold verification on the job requirement data obtained from the shipowner's terminal. Compare the minimum credit score requirement of the job requirement with the shipowner's decoupled credit score, remove invalid job requirements that do not meet the credit threshold, and output the job requirement feature vector. Perform feature concatenation processing on the initial feature vector of the crew member and the decoupled credit score of the crew member to generate a multi-dimensional feature weighted vector of the crew member. Combine the job requirement feature vector with the multi-dimensional feature weighted vector of the crew member to perform dimensional deviation calculation and weighted summation to generate a weighted comprehensive matching degree between the crew member and the job. The weighted overall matching degree is used to indicate the degree of suitability between the crew member and the current position. The higher the matching degree, the more suitable the crew member is for the current position.

2. The method according to claim 1, characterized in that, S1 includes: S11: Obtain multi-dimensional feature data of crew members and ship owners, extract the crew members' certificate level, service vessel type, service tonnage, service area and ship owner's enterprise registration information, number of vessels, vessel type and vessel safety certificate level from the multi-dimensional feature data, perform numerical mapping on the crew members' certificate level, service vessel type, service tonnage, service area and ship owner's enterprise registration information, number of vessels, vessel type and vessel safety certificate level respectively, convert each qualitative indicator into a corresponding quantitative score, and generate a crew member static feature coding sequence and a ship owner static feature coding sequence; S12: Perform time-series normalization processing on the platform login frequency, resume update cycle and training completion quantity of crew members in the multidimensional feature data, map each time-series indicator to a unified numerical range, generate a crew member dynamic feature coding sequence, and perform time-series normalization processing on the shipowner's recruitment frequency, wage payment on time rate and historical contract completion rate in the multidimensional feature data, generate a shipowner dynamic feature coding sequence. S13: Concatenate the crew member static feature encoding sequence and the crew member dynamic feature encoding sequence according to a preset feature dimension order to generate the crew member initial feature vector; and concatenate the shipowner static feature encoding sequence and the shipowner dynamic feature encoding sequence according to a preset feature dimension order to generate the shipowner initial feature vector.

3. The method according to claim 1, characterized in that, S2 includes: S21: Extract the shipowner's evaluation score, the anonymous peer score of the same crew members, and the platform's objective behavior evaluation score after the end of each contract from the historical evaluation data of the crew members in the multidimensional feature data. Then, weight and sum the shipowner's evaluation score, the anonymous peer score of the same crew members, and the platform's objective behavior evaluation score according to the preset first weight coefficient, second weight coefficient, and third weight coefficient to generate the crew member's comprehensive score for each contract. Wherein, the sum of the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient is 1; S22: Based on the time interval between the crew member's comprehensive score for each contract and the current time, determine the time decay weight coefficient corresponding to each contract, and then divide the weighted sum of the crew member's comprehensive score for each contract and the time decay weight coefficient by the sum of the time decay weight coefficients to generate the crew member's initial credit score. S23: Extract the crew's evaluation scores of the shipowner after each contract ends from the shipowner's historical evaluation data in the multidimensional feature data. Take the average of multiple crew evaluation scores within the same contract period as the shipowner's single contract comprehensive score for that contract. Calculate the shipowner's single contract comprehensive score using a weighted moving average method that is the same as the crew's initial credit score, and generate the shipowner's initial credit score.

4. The method according to claim 1, characterized in that, S3 includes: S31: Using each employment-evaluation record in the historical evaluation data of crew members and the historical evaluation data of ship owners in the multidimensional feature data as edges, and each crew member and each ship owner as nodes, construct the evaluation relationship bipartite graph; Each edge connects a crew member node and a shipowner node, and carries a first weight value and a second weight value. The first weight value is the shipowner's evaluation score of the crew member, and the second weight value is the crew member's evaluation score of the shipowner. S32: For each edge in the bipartite graph, calculate the average of the first weight values ​​of the shipowner nodes connected to the corresponding edge with respect to all historically connected crew members as the first benchmark value, calculate the average of the second weight values ​​of the crew member nodes connected to the corresponding edge with respect to all historically connected shipowners as the second benchmark value, and multiply the absolute value of the difference between the first weight value and the first benchmark value by the absolute value of the difference between the second weight value and the second benchmark value to generate the local coupling coefficient of the corresponding edge. S33: For each crew member node, sum the local coupling coefficients of all edges connected to the crew member node and divide the sum by the total number of connected edges of the crew member node to generate the global coupling strength of the crew member node; for each ship owner node, sum the local coupling coefficients of all edges connected to the ship owner node and divide the sum by the total number of connected edges of the ship owner node to generate the global coupling strength of the ship owner node.

5. The method according to claim 1, characterized in that, S4 includes: S41: Based on the comparison result of the global coupling strength and the preset coupling threshold, the initial credit scores of all shipowner nodes connected to the crew member node that is determined to have significant coupling are extracted as the initial credit score sequence of shipowners. The first weight value obtained by the crew member node on each connection edge is extracted as the shipowner evaluation score sequence corresponding to the crew member. The ratio of the covariance between the initial credit score sequence of shipowners and the evaluation score sequence of shipowners to the variance of the initial credit score sequence of shipowners is calculated to generate the shipowner factor interference coefficient. S42: Calculate the average of the initial credit scores of all shipowner nodes connected to the crew member node as the crew member's historical credit average. Obtain the global average of the initial credit scores of all shipowner nodes on the platform as the global credit average. Multiply the shipowner factor interference coefficient by the difference between the historical credit average and the global credit average. Subtract the product of the shipowner factor interference coefficient and the difference between the historical credit average and the global credit average from the initial credit score of the crew member node to generate the crew member decoupled credit score of the crew member node. S43: Based on the crew factor interference coefficient, for the shipowner node that is determined to have significant coupling, calculate the crew factor interference coefficient according to the initial credit score sequence of all crew nodes connected to the shipowner node and the crew evaluation score sequence corresponding to the shipowner. Use the crew factor interference coefficient to correct the initial credit score of the shipowner node and generate the shipowner decoupling credit score of the shipowner node.

6. The method according to claim 5, characterized in that, The formula for calculating the shipowner factor interference coefficient is as follows: in, denoted as the shipowner factor interference coefficient, used to quantitatively characterize the linear influence of shipowner credit level on the evaluation score obtained by the crew member, where d is the total number of shipowner nodes connected to the crew member node. Let the initial credit score of the shipowner at the k-th shipowner node be [the initial credit score of the shipowner]. The mean of the shipowner's initial credit score sequence. The first weight value on the k-th connecting edge is the shipowner's evaluation score of the crew. The mean of the shipowner evaluation score sequence is given.

7. The method according to any one of claims 1-6, characterized in that, S5 includes: S51: Obtain job requirement data for the current job to be matched from the shipowner terminal, extract the minimum credit score requirement field from the job requirement data, compare the minimum credit score requirement with the shipowner's decoupled credit score, if the shipowner's decoupled credit score is lower than the minimum credit score requirement, mark the job requirement as not meeting the credit threshold and remove the job requirement from the matching candidate set; if the shipowner's decoupled credit score is greater than or equal to the minimum credit score requirement, retain the job requirement. S52: The certificate level requirements, service experience requirements, and special skill requirements in the retained job requirement data are vectorized and encoded according to the same encoding rules as the initial feature vector of the crew members to generate a job requirement feature vector; S53: Perform absolute value calculation on the difference between the value of each dimension in the crew member multi-dimensional feature weighted vector and the value of the corresponding dimension in the job requirement feature vector, and subtract the absolute value of the difference of each dimension from 1 to generate the basic matching degree of each dimension. S54: Based on the job type characteristics and ship type characteristics of the current position, match the corresponding basic weight vector from the preset weight configuration scheme, and adjust the bias of the basic weight vector according to the shipowner's historical operating characteristics in the shipowner's initial feature vector to generate adaptive weights for each dimension. Multiply the basic matching degree of each dimension with the adaptive weight of the corresponding dimension and sum them to generate a weighted comprehensive matching degree.

8. A crew member and job matching evaluation system based on multi-dimensional feature weighting, characterized in that, The system includes: The multidimensional feature encoding module is used to acquire multidimensional feature data of crew members and ship owners, and to perform vectorization encoding on the multidimensional feature data to generate initial feature vectors for crew members and initial feature vectors for ship owners. The main credit calculation module is used to calculate the crew member's single contract comprehensive score by weighted summing of the multi-party evaluation scores in the historical evaluation data in the multi-dimensional feature data, and to generate the crew member's initial credit score and the shipowner's initial credit score by weighted moving average of the crew member's single contract comprehensive score. The bipartite graph coupling strength solution module is used to construct a bipartite graph of the evaluation relationship between crew members and ship owners based on the historical evaluation data. It performs a deviation product operation on the evaluation score on each edge of the bipartite graph and the initial credit scores of the crew members and the ship owners at the corresponding two ends of the node to generate the global coupling strength between each crew member node and each ship owner node. The credit score covariance decoupling module is used to filter out the crew credit scores and shipowner credit scores that have significant coupling based on the global coupling strength, and to perform covariance decoupling correction processing on the crew credit scores and shipowner credit scores that have significant coupling, thereby generating crew decoupled credit scores and shipowner decoupled credit scores. The comprehensive matching calculation module for personnel and positions is used to perform credit threshold verification on the job requirement data obtained from the shipowner terminal based on the shipowner's decoupled credit score. It compares the minimum credit score requirement of the job requirement with the shipowner's decoupled credit score, eliminates invalid job requirements that do not meet the credit threshold, and outputs a job requirement feature vector. It performs feature concatenation processing on the initial feature vector of the crew member and the crew member's decoupled credit score to generate a multi-dimensional feature weighted vector of the crew member. It combines the job requirement feature vector with the multi-dimensional feature weighted vector of the crew member to perform dimensional deviation calculation and weighted summation to generate a weighted comprehensive matching degree between the crew member and the job. The weighted overall matching degree is used to indicate the degree of suitability between the crew member and the current position. The higher the matching degree, the more suitable the crew member is for the current position.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images