AI-based human resource talent rapid matching system

Abstract

The invention discloses an AI-based human resource talent rapid matching system, which comprises a feature acquisition module, a standardized splicing module, a matrix pre-construction module and an anchoring module, and is used for anchoring a most similar historical object in a historical entry matrix of a historical entry according to a job hunting feature vector of a current job hunting person; the execution module is used for executing an in-job matching program of the current job seeker according to the most similar historical object; the matching step comprises the following steps: extracting a real work score from a two-tuple sample of the most similar historical object; inheriting the extracted real work score as a work valuation score of the current job seeker; if the work estimation score is greater than the threshold value, judging that the current job seeker is completely matched, otherwise, judging that the current job seeker is not completely matched; according to the method, the historical object most similar to the features of the current job seeker is anchored in the historical job entry matrix, cross-time-sequence feature inheritance reasoning based on the behavior similar body is constructed, and job entry adaptation evaluation and matching judgment of the current job seeker are completed.

Description

Technical Field

[0001] The present invention relates to a matching system, specifically an AI-based rapid matching system for human resources talents. Background Art

[0002] With the informatization and intelligent development of human resource management, more and more enterprises begin to adopt candidate screening and job matching technologies based on artificial intelligence algorithms. The existing technologies generally adopt a supervised learning-based AI matching scoring model, which extracts multi-dimensional features of job seekers (such as education background, skills, experience, etc.), constructs a prediction model in combination with job requirements, and scores or ranks the suitability of candidates. The patent document with the patent publication number CN118333591A discloses a scheduling method and device for human resources based on dynamic optimization, which realizes the best matching of job requirements and human resources, and solves the problems of strong subjectivity and low matching accuracy existing in traditional matching methods.

[0003] However, in the existing technologies, the intelligent matching of human resources mainly relies on supervised learning models. By constructing a prediction model for a large-scale sample and reasoning about the suitability of job seekers, such methods have the following technical defects:

[0004] It is necessary to construct a prediction model for all candidate objects in the job to be matched and reason about the suitability of job seekers one by one. It depends on a large number of training samples and complex model structures, lacks the ability of time dynamic adaptation, and is not conducive to rapid screening for real-time recruitment tasks. And it ignores the influence of time factors on the judgment of person-job matching, and cannot make full use of the behavior trajectories of historical onboarding samples adjacent in time as a matching reference, resulting in a lack of real-time prediction. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technologies, the present invention provides an AI-based rapid matching system for human resources talents, which solves the technical problems raised in the background art by inheriting the work valuation scores of the most similar historical objects.

[0006] To achieve the above purposes, the present invention is realized through the following technical solutions:

[0007] An AI-based rapid matching system for human resources talents, the matching system includes:

[0008] A feature acquisition module, used to acquire several job-seeking features of the current job seeker;

[0009] A standardization splicing module, used to standardize several job-seeking features and splice the standardized several job-seeking features to generate a job-seeking feature vector of the current job seeker;

[0010] A matrix pre-construction module, used to pre-construct a historical onboarding matrix of historical onboarding employees;

[0011] Among them, the historical employment matrix is characterized by binary group samples of multiple historical employees in different employment time windows. Each binary group sample includes job-hunting characteristics and corresponding real job ratings.

[0012] The anchoring module is used to anchor the most similar historical object in the historical employment matrix of historical employees according to the job-hunting characteristic vector of the current job applicant.

[0013] The execution module is used to execute the employment matching program of the current job applicant according to the most similar historical object.

[0014] The matching steps of the employment matching program are as follows:

[0015] Extract the real job rating from the binary group sample of the most similar historical object.

[0016] Inherit the extracted real job rating as the job valuation rating of the current job applicant.

[0017] If the job valuation rating is greater than the threshold, it is determined that the current job applicant is fully matched; otherwise, it is not fully matched.

[0018] In some specific embodiments, the matrix pre-construction module is specifically used for:

[0019] S3-1. Obtain several job-hunting characteristics of historical employees before employment within the employment time window.

[0020] S3-2. Construct the binary group sample of the historical employee according to several historical job-hunting characteristics of the historical employee before employment within the employment time window.

[0021] S3-3. Obtain the binary group samples of N historical job applicants.

[0022] S3-4. Based on the binary group samples of N historical job applicants, construct the matrix row vectors within the employment time window. Among them, the matrix row vectors are characterized by the job-hunting characteristic sequences of historical job applicants composed of binary group samples.

[0023] S3-5. Obtain M matrix row vectors within the employment time window. Each matrix row vector within the employment time window is marked with an employment timestamp.

[0024] S3-6. Align the matrix row vectors within the employment time window with the M matrix row vectors within the employment time window to obtain the historical employment matrix of N*M historical employees.

[0025] In some specific embodiments, the generation steps of the binary group sample of the historical employee include:

[0026] S3-2-1. Perform feature standardization on several job hunting characteristics of historical recruits within the employment time window before employment;

[0027] S3-2-2. Concatenate the standardized job hunting characteristics to generate the job hunting feature vector of the historical recruit;

[0028] S3-2-3. Collect the true job performance ratings of historical recruits within a specified time period;

[0029] S3-2-4. Pair the job hunting feature vector of the historical recruit with the true job performance rating to generate the binary sample of the historical recruit.

[0030] In some specific embodiments, the steps for constructing the matrix row vector within the employment time window include:

[0031] S3-4-1. Assign sequential job hunting numbers to N historical job seekers;

[0032] S3-4-2. Sort the binary samples of N historical job seekers according to the sequential job hunting numbers to generate a sequence of binary samples of N historical job seekers;

[0033] S3-4-3. Define the sequence of binary samples of N historical job seekers as the matrix row vector within the employment time window; where each element of the matrix row vector is a binary sample.

[0034] In some specific embodiments, the anchoring module is specifically used for:

[0035] S4-1. Calculate the similarity between the job hunting feature vector of the current job seeker and each job hunting feature vector in the historical employment matrix of historical recruits to obtain N*M similarities;

[0036] S4-2. Calculate the time weight of each matrix row vector in the historical employment matrix;

[0037] S4-3. Determine the most matching similarity among several historical job seekers according to the N*M similarities and the time weight of each matrix row vector;

[0038] S4-4. Define the historical job seeker corresponding to the most matching similarity as the most similar historical object.

[0039] In some specific embodiments, the steps for calculating the time weight include:

[0040] S4-2-1. Extract the employment timestamp of each matrix row vector from the historical employment matrix of historical recruits;

[0041] S4-2-2. Calculate the time interval between the employment timestamp and the current job hunting timestamp;

[0042] S4-2-3. After performing an inverse calculation on the time interval, generate the time weight.

[0043] In some specific embodiments, the determining step of the most matching similarity includes:

[0044] S4-3-1. Extract the time weight of each matrix row vector in the historical employment matrix;

[0045] S4-3-2. Based on the time weight of each matrix row vector and the N similarities in the matrix row vector, construct a first similarity ordered sequence;

[0046] S4-3-3. Obtain the first similarity ordered sequences within M employment time windows;

[0047] S4-3-4. From the first similarity ordered sequences within M employment time windows, construct a second similarity ordered sequence;

[0048] S4-3-5. Determine the most matching similarity from the second similarity ordered sequence.

[0049] In some specific embodiments, the constructing step of the first similarity ordered sequence includes:

[0050] S4-3-2-1. Calculate the product of the time weight of each matrix row vector and the N similarities in the matrix row vector to generate N first similarities;

[0051] S4-3-2-2. Sort the N first similarities to generate the corresponding first similarity ordered sequence within the employment time window; wherein, the first similarity ordered sequence represents the ordered arrangement of the time-weighted similarities within the employment time window.

[0052] In some specific embodiments, the constructing step of the second similarity ordered sequence includes:

[0053] S4-3-4-1. From the first similarity ordered sequences within M employment time windows, select the maximum first similarity in each ordered sequence;

[0054] S4-3-4-2. Define the maximum first similarity in each ordered sequence as the second similarity to obtain M second similarities;

[0055] S4-3-4-3. Sort the M second similarities to generate the second similarity ordered sequence; wherein, the second similarity ordered sequence represents the ordered sequence among the maximum first similarities of all employment time windows.

[0056] In some of these specific embodiments, determining the most matching similarity from the second similarity ordered sequence includes:

[0057] S4-3-5-1. Select the maximum second similarity for the second similarity ordered sequence;

[0058] S4-3-5-2. If the number of the maximum second similarities is greater than 1, mark the maximum second similarity corresponding to the employment timestamp closest to the current job application timestamp;

[0059] S4-3-5-3. Define the marked maximum second similarity as the most matching similarity.

[0060] The present invention provides an AI-based rapid matching system for human resources talents, having the following beneficial effects:

[0061] By constructing a historical employment information matrix covering multiple employment time windows and combining dynamic weighted similarity calculation based on time weights, the present invention forms a strategy of comprehensive weighting based on time sensitivity and similarity, which can adjust the contribution intensity of the binary group samples to the current job applicant during the matching process, so as to adapt to the recruitment rhythm and the time fluctuations in the talent market. It significantly improves the time adaptability of the matching results and effectively solves the defects of ignoring the time impact and having a static and rigid matching strategy in the prior art.

[0062] Furthermore, by anchoring the historical object most similar to the current job applicant's characteristics in the historical employment matrix and directly inheriting its real work score in the actual position as the work valuation score of the current job applicant, the present invention constructs a cross-temporal feature inheritance inference based on behavioral similarity bodies to complete the employment adaptation evaluation and matching judgment of the current job applicant. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a structural block diagram of an AI-based rapid matching system for human resources talents of the present invention;

[0064] Figure 2 It is a schematic diagram of the matching process of an AI-based rapid matching system for human resources talents of the present invention;

[0065] Figure 3 It is a schematic diagram of the definition process of the most similar historical object of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0067] Embodiment 1: Please refer to Figures 1 to 3 , the present invention provides an AI-based rapid human resource talent matching system, and the matching system includes:

[0068] A feature acquisition module for acquiring a number of job-seeking features of the current job seeker;

[0069] The job-seeking features include but are not limited to:

[0070] Gender: The gender information of the job seeker (such as male, female, other), which is used for matching analysis of specific positions or industries.

[0071] Age range: The age or age group of the job seeker (such as 18-25 years old, 26-35 years old, etc.).

[0072] Physical sign parameters: including height, weight, physical health status, etc., which are applicable to some specific industries (such as sports, medical, etc.) or specific positions.

[0073] Educational attainment range: The highest educational attainment information of the job seeker (such as undergraduate, master, doctor, etc.), which is used to evaluate whether the job seeker meets the basic educational requirements of the position.

[0074] Skill types and their skill levels: The skill types possessed by the job seeker (such as programming, marketing, project management, etc.), and the proficiency levels of these skills (such as junior, intermediate, senior), which are used to evaluate whether the job seeker has the technical capabilities required for the position.

[0075] Work experience: The number of years of work experience of the job seeker and the relevant positions or industries he has engaged in, which is crucial for matching positions with high experience requirements.

[0076] Geographical location: The place of residence of the job seeker or the geographical area where he is willing to work, which is particularly important for positions with geographical restrictions.

[0077] Language ability: The languages mastered by the job seeker and their fluency levels, which are particularly applicable to multinational companies or multilingual positions.

[0078] Job-seeking intention: The preferences of the job seeker for job positions (such as full-time, part-time, remote work, etc.), which helps to better match his job expectations.

[0079] A standardization splicing module is used to standardize a number of job-seeking features and then splice the standardized job-seeking features to generate a job-seeking feature vector for the current job applicant.

[0080] The standardization process can perform different standardization preprocessings based on the feature attributes of the job-seeking features. Among them, the feature splicing includes sequentially splicing each standardized job-seeking feature vector together to generate a comprehensive job-seeking feature vector. Thus, the dimensions of all job-seeking features will be retained, forming a more comprehensive representation of the job applicant.

[0081] A matrix pre-construction module is used to pre-construct a historical employment matrix of historical employees.

[0082] Among them, the historical employment matrix is characterized as a binary sample of multiple historical employees in different employment time windows. Each binary sample includes a job-seeking feature and the corresponding true job score.

[0083] In this embodiment, the matrix pre-construction module is specifically used for:

[0084] S3-1: Obtain a number of job-seeking features of historical employees before employment within the employment time window.

[0085] S3-2: Construct the binary sample of the historical employee according to a number of historical job-seeking features of the historical employee before employment within the employment time window.

[0086] S3-3: Obtain the binary samples of N historical job applicants.

[0087] S3-4: Based on the binary samples of N historical job applicants, construct a matrix row vector within the employment time window. The matrix row vector is characterized as a job-seeking feature sequence of historical job applicants composed of binary samples.

[0088] S3-5: Obtain M matrix row vectors within the employment time window. Each matrix row vector within the employment time window is marked with an employment timestamp.

[0089] S3-6: Align the matrix row vectors within the employment time window with the M matrix row vectors within the employment time window to obtain a historical employment matrix of N*M historical employees.

[0090] In this embodiment, the system forms a structured historical onboarding matrix by constructing historical onboarding information covering multiple onboarding time windows, realizing the temporal modeling of historical human matching data. The elements of this matrix are "binary tuple samples", that is, the job hunting characteristics of each historical job seeker at a certain onboarding time point and their corresponding true job scores. The binary tuple samples record the correspondence between the input and the matching results. By dividing the binary tuple samples of multiple historical onboarding employees according to time windows, matrix row vectors are formed, and then a unified and aligned matrix representation is established under multiple time windows, enabling the capture of human matching trends in different historical stages simultaneously. This historical onboarding matrix not only has the ability to compare individual differences horizontally (between different historical objects), but also has the ability to express temporal evolution vertically (for the same object in different time windows), providing high-quality binary tuple sample support with time tags for subsequent matching anchoring, significantly improving the system's job seeker matching ability in multiple periods and multiple positions.

[0091] Exemplarily, in this embodiment, the steps for generating the binary tuple samples of the historical onboarding employees include:

[0092] S3-2-1. Perform feature standardization on several job hunting characteristics of the historical onboarding employees before onboarding within the onboarding time window;

[0093] S3-2-2. Concatenate the standardized several job hunting characteristics to generate the job hunting feature vector of the historical onboarding employee;

[0094] S3-2-3. Collect the true job scores of the historical onboarding employees within a specified duration;

[0095] S3-2-4. Pair the job hunting feature vector of the historical onboarding employee with the true job score to generate the binary tuple sample of the historical onboarding employee.

[0096] In this embodiment, by collecting the true job scores of the historical onboarding employees within a certain period after onboarding, a one-to-one correspondence between the input features and the target results is constructed, and finally the two are paired to generate binary tuple samples. The binary tuple samples have a data structure of behavioral prior and result feedback, and can reflect the causal relationship between job hunting characteristics and job competence.

[0097] Exemplarily, in this embodiment, the steps for constructing the matrix row vectors within the onboarding time window include:

[0098] S3-4-1. Assign sequential job hunting numbers to N historical job seekers;

[0099] Among them, the sequential job application number can be used to identify the sequential position of each historical job applicant in the historical employment matrix, and its generation method includes but is not limited to: combining the applicant's unique ID with their initial employment timestamp, and then sorting them in chronological order of employment to generate a continuously increasing number sequence.

[0100] S3-4-2. Sort the binary tuple samples of N historical job applicants according to the sequential job application number to generate a sequence of binary tuple samples of N historical job applicants;

[0101] S3-4-3. Define the sequence of binary tuple samples of N historical job applicants as the matrix row vectors within the employment time window; where each matrix row vector element is a binary tuple sample.

[0102] In this embodiment, by assigning a sequential job application number with a chronological logic to each historical job applicant, the orderly organization of binary tuple samples in employment information is achieved. The sequential job application number can be generated based on the combination of the applicant's unique identity and their first employment timestamp, and sorted in chronological order to ensure that each historical job applicant has a traceable sequential position in the repository.

[0103] The reference Figure 1 , the matching system further includes:

[0104] An anchoring module for anchoring the most similar historical object in the historical employment matrix of historical employees according to the job application feature vector of the current job applicant;

[0105] An execution module for executing the employment matching program of the current job applicant according to the most similar historical object;

[0106] The matching steps of the employment matching program are as follows:

[0107] Extract the real job score from the binary tuple samples of the most similar historical object;

[0108] Inherit the extracted real job score as the job valuation score of the current job applicant;

[0109] If the job valuation score is greater than the threshold, it is determined that the current job applicant is a complete match; otherwise, it is an incomplete match.

[0110] In this embodiment, by structuring the multi-dimensional job application characteristics of the current job applicant, a job application feature vector is constructed and used as the basic input to match with the historical employment matrix of historical employees.

[0111] Specifically, in the form of binary sample groups, the information matrix records the pre-employment characteristic information of a large number of historical job seekers and their actual job performance, and has typical human resource sample value. By anchoring the most similar historical employed objects in this matrix and inheriting their actual true job scores as the predicted performance values of current job seekers, the system can achieve rapid ability valuation of current job seekers without performing additional regression reasoning. Finally, based on the comparison result between this valuation and the set threshold, a classification judgment of "perfect match" or not is given, thus constituting a human resource matching method based on inheritance of similar human trajectories.

[0112] Embodiment 2: Refer to Figures 1 to 3 The technical solution of this Embodiment 2 is different from that of Embodiment 1 in that the specific anchoring steps of the anchoring module described in Embodiment 1 are disclosed; the anchoring steps include:

[0113] S4-1. Calculate the similarity between the job-seeking feature vector of the current job seeker and each job-seeking feature vector in the historical employment matrix of historical employed persons, and obtain N*M similarities;

[0114] Among them, the cosine similarity or Euclidean distance is used for similarity calculation. The cosine similarity can ignore the absolute size of features and only consider the relationship between features, and is particularly suitable for the matching between job seeker skills (such as skill matching) and job requirements. Moreover, the calculation is simple, suitable for vectorized features, and no additional processing of the scale between features is required. The Euclidean distance is suitable for similarity calculation between numerical features (such as age, years of work experience, weight, education level, etc.). It calculates the similarity by measuring the straight-line distance between two points.

[0115] S4-2. Calculate the time weight of each matrix row vector in the historical employment matrix;

[0116] S4-3. Determine the most matching similarity among several historical job seekers according to the N*M similarities and the time weight of each matrix row vector;

[0117] S4-4. Define the historical job seeker corresponding to the most matching similarity as the most similar historical object.

[0118] In this embodiment, a full-scale similarity calculation is performed between the job application feature vector of the current job seeker and all the historical job application feature vectors in the historical employment matrix, constructing a similarity set covering all binary tuple samples and time windows. Meanwhile, a time weight mechanism is introduced to adjust the matrix row vectors, such that binary tuple samples closer to the current job application time have higher matching weights. Based on the combined sorting result of the similarity scores and time weights, the system dynamically anchors the binary tuple sample with the optimal match as the "most similar historical object" of the current job seeker. This anchoring strategy combines two-dimensional judgments of horizontal feature similarity and vertical time correlation, and can identify the historical employee with the most reference value for the current job seeker among the dense binary tuple samples.

[0119] Exemplarily, in this embodiment, the calculation steps of the time weight include:

[0120] S4-2-1: Extract the employment timestamps of each matrix row vector from the historical employment matrix of historical employees;

[0121] S4-2-2: Calculate the time interval between the employment timestamp and the current job application timestamp;

[0122] S4-2-3: After performing an inverse calculation on the time interval, generate the time weight;

[0123] Among them, the inverse calculation generates the time weight by calculating the reciprocal of the time interval. The specific calculation formula is:

[0124]

[0125] Among them, W t represents the time weight, ΔT is the time interval between the historical employment timestamp and the current job application timestamp, and ∈ is a small constant used to avoid division by zero errors. Exemplarily, in this embodiment, the role of the inverse calculation is to ensure that historical data with shorter time intervals has a greater impact on the match of the current job seeker, while historical data with longer time intervals has a smaller impact on the match.

[0126] In this embodiment, for each timestamp of employment corresponding to each row vector in the historical employment matrix, the time interval between it and the current job application timestamp is calculated, and an inverse transformation is performed based on this interval to generate the corresponding time weight. By introducing this time weight, the contribution degree of different historical data to the current matching result can be dynamically adjusted when performing comprehensive similarity calculation. Specifically, the closer the historical employment sample is to the current job application time point, the greater its time weight, so it obtains a higher matching influence in the process of anchoring the most similar historical object; conversely, historical samples that are far in time will automatically weaken their influence due to lower weights. This time decay mechanism effectively avoids the adaptation deviation problem caused by the too long time span of historical data, makes the matching strategy more time-sensitive, and thus improves the accuracy of the system in the human resource matching scenario.

[0127] Exemplarily, in this embodiment, the steps for determining the most matching similarity include:

[0128] S4-3-1. Extract the time weight of each matrix row vector in the historical employment matrix;

[0129] S4-3-2. Construct a first similarity ordered sequence according to the time weight of each matrix row vector and the N similarities in the matrix row vector;

[0130] S4-3-3. Obtain the first similarity ordered sequences within M employment time windows;

[0131] S4-3-4. Construct a second similarity ordered sequence from the first similarity ordered sequences within M employment time windows;

[0132] S4-3-5. Determine the most matching similarity from the second similarity ordered sequence.

[0133] In this embodiment, according to the matching relationship between the current job applicant and the historical employee, considering both similarity and time factors, an optimal similarity screening method based on weighted matching sorting is proposed. Specifically, this method first extracts the time weight of each matrix row vector in the historical employment matrix, and then performs weighted fusion with the similarity value of each historical object in this row to form a time-sensitive first similarity ordered sequence.

[0134] After the system generates this sequence in each time window respectively, it gradually summarizes the representative optimal values in each time window, constructs a second similarity ordered sequence across time windows, and finally determines the globally optimal matching object from this sequence.

[0135] Through the screening method, a layer-by-layer screening from local matching to global time weight adjustment and then to the final most similar historical object is achieved.

[0136] Exemplarily, in this embodiment, the steps for constructing the first similarity ordered sequence include:

[0137] S4-3-2-1. Calculate the product of the time weight of each matrix row vector and the N similarities in the matrix row vector to generate N first similarities;

[0138] S4-3-2-2. Sort the N first similarities to generate the first similarity ordered sequence within the corresponding employment time window; wherein, the first similarity ordered sequence represents the ordered arrangement of the time-weighted similarities within the employment time window.

[0139] In this embodiment, for each matrix row vector in the historical employment matrix, first extract its time weight value, and perform weighted calculation item by item on it and the N similarities generated by the current job applicant to generate a set of first similarities after time weighting. This set is fused by multiplication to combine the job-seeking feature similarity and time approximation, so that historical objects closer to the current job-seeking time occupy a higher weight in the calculation. Subsequently, sort this set to form the first similarity ordered sequence within the employment time window, so that in each window, the historical object with the highest matching degree can be quickly located.

[0140] Exemplarily, in this embodiment, the steps for constructing the second similarity ordered sequence include:

[0141] S4-3-4-1. Select the maximum first similarity in each ordered sequence from the first similarity ordered sequences within M employment time windows;

[0142] S4-3-4-2. Define the maximum first similarity in each ordered sequence as the second similarity to obtain M second similarities;

[0143] S4-3-4-3. Sort the M second similarities to generate the second similarity ordered sequence; wherein, the second similarity ordered sequence represents the ordered sequence among the maximum first similarities of all employment time windows.

[0144] In this embodiment, based on the first similarity ordered sequences within the M time windows already obtained by the system, further extract the maximum value in each sequence to form the second similarity ordered sequence. Each maximum first similarity is regarded as the similarity between the historical object with the highest matching degree under the current time window and the current job applicant, thus forming a set of binary tuple samples representing the optimal matching quality in each time stage.

[0145] Exemplarily, in this embodiment, determining the most matching similarity from the second similarity ordered sequence includes:

[0146] S4-3-5-1. Select the maximum second similarity from the second similarity ordered sequence;

[0147] S4-3-5-2. If the number of the maximum second similarities is greater than 1, mark the maximum second similarity corresponding to the entry timestamp closest to the current job application timestamp;

[0148] S4-3-5-3. Define the marked maximum second similarity as the most matching similarity.

[0149] In this embodiment, the system preferentially selects the similarity with the highest matching value from the second similarity ordered sequence as the historical path sample that the current job seeker is most likely to refer to. If there are multiple candidates with the maximum second similarity, a time priority strategy is further introduced to mark the binary tuple sample closest to the current job application timestamp from them to ensure the practical availability of the matching object in the time dimension. Finally, the selected similarity and its corresponding historical object are defined as the most matching similarity and the most similar historical object.

[0150] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means.

[0151] The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or a data center containing one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0152] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division of an underwater topographic change analysis system and method for waterways. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical, mechanical, or other form.

[0153] As described above, the above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, and all should be covered within the protection scope of the present application.

Claims

1. An AI-based rapid matching system for human resources talents, characterized in that, The matching system includes: A feature acquisition module, which is used to acquire several job-seeking features of the current job applicant; A standardization splicing module, which is used to standardize several job-seeking features and then splice the standardized job-seeking features to generate a job-seeking feature vector of the current job applicant; A matrix pre-construction module, which is used to pre-construct a historical employment matrix of historical employees; Among them, the historical employment matrix is characterized by binary group samples of multiple historical employees in different employment time windows, where each binary group sample includes a job-seeking feature and a corresponding real job score; An anchoring module, which is used to anchor the most similar historical object in the historical employment matrix of historical employees according to the job-seeking feature vector of the current job applicant; An execution module, which is used to execute the employment matching program of the current job applicant according to the most similar historical object; The matching steps of the employment matching program are: Extract the real job score from the binary group samples of the most similar historical object; Inherit the extracted real job score as the job valuation score of the current job applicant; If the job valuation score is greater than the threshold, it is determined that the current job applicant is fully matched, otherwise it is not fully matched.

2. The rapid matching system for human resources talents based on AI according to claim 1, characterized in that Specifically, the matrix pre-construction module is used for: S3-1. Acquire several job-seeking features of historical employees before employment within the employment time window; S3-2. Construct the binary group samples of the historical employees according to several historical job-seeking features of historical employees before employment within the employment time window; S3-3. Acquire binary group samples of N historical job applicants; S3-4. Based on the binary group samples of N historical job applicants, construct matrix row vectors within the employment time window; where the matrix row vectors are characterized by a sequence of job-seeking features of historical job applicants composed of binary group samples; S3-5. Acquire M matrix row vectors within the employment time window; where each matrix row vector within the employment time window is marked with an employment timestamp; S3-6. Align the matrix row vectors within the employment time window with the M matrix row vectors within the employment time window to obtain a historical employment matrix of N*M historical employees.

3. The rapid matching system for human resources talents based on AI according to claim 2, characterized in that, The generation steps of the binary group samples of the historical employees include: S3-2-1. Perform feature standardization on several job-seeking features of historical employees before employment within the employment time window; S3-2-2. Splice the standardized job-seeking features to generate a job-seeking feature vector of the historical employee; S3-2-3. Collect the real job score of the historical employee within a specified duration; S3-2-4. Pair the job-seeking feature vector of the historical employee with the real job score to generate the binary group sample of the historical employee.

4. An AI-based rapid human resource talent matching system according to claim 3, characterized in that, The construction steps of the matrix row vectors within the employment time window include: S3-4-1. Assign sequential job-seeking numbers to N historical job applicants; S3-4-2. Sort the binary group samples of N historical job applicants according to the sequential job-seeking numbers to generate a sequence of binary group samples of N historical job applicants; S3-4-3. Define the sequence of binary group samples of N historical job applicants as the matrix row vectors within the employment time window; where each matrix row vector element is a binary group sample.

5. The rapid matching system for human resources talents based on AI according to claim 1 is characterized in that, The anchoring module is specifically used for: S4-1. Calculate the similarity between the job-seeking feature vector of the current job seeker and each job-seeking feature vector in the historical employment matrix of historical employees, obtaining N*M similarities; S4-2. Calculate the time weight of each matrix row vector in the historical employment matrix; S4-3. Determine the most matching similarity among several historical job seekers according to the N*M similarities and the time weight of each matrix row vector; S4-4. Define the historical job seeker corresponding to the most matching similarity as the most similar historical object.

6. The rapid matching system for human resources talents based on AI according to claim 5, characterized in that The calculation steps of the time weight include: S4-2-1. Extract the employment timestamp of each matrix row vector from the historical employment matrix of historical employees; S4-2-2. Calculate the time interval between the employment timestamp and the current job-seeking timestamp; S4-2-3. After inverse calculation of the time interval, generate the time weight.

7. The AI-based rapid human resource talent matching system according to claim 6, characterized in that, The determination steps of the most matching similarity include: S4-3-1. Extract the time weight of each matrix row vector in the historical employment matrix; S4-3-2. Construct a first similarity ordered sequence according to the time weight of each matrix row vector and the N similarities in the matrix row vector; S4-3-3. Obtain the first similarity ordered sequences within M employment time windows; S4-3-4. Construct a second similarity ordered sequence from the first similarity ordered sequences within M employment time windows; S4-3-5. Determine the most matching similarity from the second similarity ordered sequence.

8. The rapid matching system for human resources talents based on AI according to claim 7, characterized in that, The construction steps of the first similarity ordered sequence include: S4-3-2-1. Calculate the product of the time weight of each matrix row vector and the N similarities in the matrix row vector, generating N first similarities; S4-3-2-2. Sort the N first similarities to generate the first similarity ordered sequence within the corresponding employment time window; wherein, the first similarity ordered sequence represents the ordered arrangement of the time-weighted similarities within the employment time window.

9. The rapid matching system for human resources talents based on AI according to claim 8, characterized in that, The construction steps of the second similarity ordered sequence include: S4-3-4-1. Select the maximum first similarity in each ordered sequence from the first similarity ordered sequences within M employment time windows; S4-3-4-2. Define the maximum first similarity in each ordered sequence as the second similarity, obtaining M second similarities; S4-3-4-3. Sort the M second similarities to generate the second similarity ordered sequence; wherein, the second similarity ordered sequence represents the ordered sequence among the maximum first similarities of all employment time windows.

10. A rapid matching system for human resources talents based on AI according to claim 9, characterized in that, Determining the most matching similarity from the second similarity ordered sequence includes: S4-3-5-1. Select the maximum second similarity from the second similarity ordered sequence; S4-3-5-2. If the number of the maximum second similarities is greater than 1, mark the maximum second similarity corresponding to the employment timestamp closest to the current job-seeking timestamp; S4-3-5-3. Define the marked maximum second similarity as the most matching similarity.

Images