Machine learning based multi-dimensional feature and post image matching recommendation system

Abstract

This invention discloses a recommendation system based on machine learning and multi-dimensional feature matching with job profiles, belonging to the field of profile matching technology. This invention constructs a multi-level, three-dimensional student feature evaluation system, significantly improving the objectivity, accuracy, and dimensionality of student ability feature expression. This enables the system to comprehensively capture students' explicit work abilities and implicit professional qualities, forming a more realistic and accurate ability profile. It avoids the subjective bias and information distortion problems of traditional self-assessment questionnaires. This invention can deeply mine behavioral characteristics and cognitive patterns during task execution, significantly improving the accuracy of ability assessment and the suitability of job matching. Simultaneously, this invention implements a personalized job assessment strategy and, based on a geometric modeling method of three-dimensional feature cones, transforms abstract multi-dimensional ability features into a visualized and quantifiable spatial profile, improving the acceptance of recommendation results and the employment success rate.

Description

Technical Field

[0001] This invention relates to the field of profile matching technology, specifically to a recommendation system that matches multidimensional features based on machine learning with job profiles. Background Technology

[0002] As a core technology platform connecting job seekers with corporate job requirements, the job recommendation system's matching accuracy and personalization directly affect students' career development paths and companies' talent selection efficiency. Therefore, it is essential to build a machine learning-based multi-dimensional feature intelligent analysis and job profile accurate matching system for in-depth ability assessment and scientific recommendation.

[0003] In existing technologies, most job recommendation systems rely on traditional resume information and standardized assessment questionnaires, which cannot cover students' dynamic ability performance and multi-level feature recognition in real work scenarios. Although some existing technologies have introduced personality test scales and skills assessment tools, due to their limited data collection dimensions and high dependence on students' self-evaluation, it is difficult to achieve an objective judgment on students' actual work ability, task execution efficiency and implicit professional qualities. When faced with diverse job types, it is easy to have problems such as recommendation results not matching students' actual abilities and low job suitability.

[0004] Existing technologies lack the ability to jointly collect, deeply analyze, and intelligently match multimodal behavioral data such as students' task completion paths, operational proficiency, attention allocation patterns, and decision-making hesitation levels in simulated work scenarios. They also lack a mechanism to identify the differentiated impact of different personality traits on different job types, making it impossible to effectively achieve individualized feature parameter tuning and personalized assessment. Furthermore, existing systems suffer from drawbacks such as difficulty in verifying data authenticity, long feedback cycles, insufficient feature mining depth, and low model optimization efficiency. This results in a significant gap between recommended results and students' actual job competence. In addition, they lack adaptive strategies for the feature requirements of different job types and cannot dynamically adjust feature weights and matching mechanisms according to the differentiated capability requirements of different job scenarios such as product, design, development, testing, operation and maintenance, and maintenance. Summary of the Invention

[0005] The purpose of this invention is to provide a recommendation system that matches multidimensional features with job profiles based on machine learning, thereby solving the problems existing in the background technology.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a recommendation system based on machine learning for matching multidimensional features with job profiles, including: a multidimensional feature acquisition module, used to conduct a simulated assessment of target students in scientific research enterprises, thereby acquiring the explicit feature data and implicit feature data of the target students.

[0007] The student profile fitting module is used to perform multidimensional feature analysis based on the explicit and implicit feature data of the target student, thereby obtaining the explicit feature set and implicit feature set of the target student. Then, vector analysis is performed on the explicit feature set and implicit feature set of the target student to obtain the target feature profile of the target student.

[0008] The job feature and student profile matching module is used to obtain each preset job type from the local database, and calculate the matching recommendation index between the target student and each preset job type based on the target student's target feature profile and a hybrid recommendation analysis method.

[0009] The job recommendation processing module is used to generate a job recommendation list for the target student based on the matching recommendation index between the target student and each preset job type, and send it to the target student for visual recommendation.

[0010] The beneficial effects of this invention are as follows: This invention constructs a multi-level, three-dimensional student characteristic assessment system, which significantly improves the objectivity, accuracy, and dimensionality of student ability characteristic expression. The system can comprehensively capture students' explicit work abilities and implicit professional qualities, forming a more realistic and accurate ability profile. This avoids the subjective bias and information distortion problems of traditional self-assessment questionnaires, providing a scientific data foundation for subsequent job matching. This invention can deeply mine behavioral characteristics and cognitive patterns during task execution, significantly improving the accuracy of ability assessment and the suitability of job matching. Simultaneously, this invention implements a personalized job assessment strategy and, based on the geometric modeling method of a three-dimensional feature cone, transforms abstract multi-dimensional ability characteristics into a visualized and quantifiable spatial profile, achieving scientific and precise person-job matching. Furthermore, it combines students' job preference coefficients for comprehensive recommendations, balancing objective ability suitability and subjective willingness guidance, thereby improving the acceptance of recommendation results and employment success rate. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the system modules of the present invention.

[0013] Figure 2 This is the special bottom triangle of the present invention.

[0014] Attached label: 1. Adjust the length value represented by the proficiency assessment coefficient; 2. Adjust the length value represented by the attention concentration coefficient; 3. Adjust the length value represented by the operation hesitation coefficient; 4. The center point of the special bottom triangle. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Reference Figure 1 As shown, the present invention provides a recommendation system based on machine learning for matching multidimensional features with job profiles, including: a multidimensional feature acquisition module, a student profile fitting module, a job feature and student profile matching module, and a job recommendation processing module.

[0017] It should be noted that the multidimensional feature acquisition module is connected to the student profile fitting module, the student profile fitting module is connected to the job feature and student profile matching module, and the job feature and student profile matching module is connected to the job recommendation processing module.

[0018] The multidimensional feature acquisition module is used to conduct simulated assessments of target students in scientific research enterprises, thereby obtaining explicit and implicit feature data of the target students.

[0019] In a specific embodiment of the present invention, the method for obtaining the explicit and implicit characteristic data of the target student is as follows: obtain the simulated test scenarios of the research enterprise from the local database, obtain the task completion time and task completion quality score of the target student in each simulated test scenario, and use them as the explicit characteristic data of the target student.

[0020] It should be noted that the local database is used to store various simulated test scenarios of scientific research enterprises, average task completion time, average task completion quality score, interface behavior analysis model, proficiency evaluation coefficient tuning values ​​corresponding to each conversion purpose, screen image analysis model, screen images of target students at each monitoring time point in each simulated test scenario, final draft text of target students in each simulated test scenario corresponding to each deleted text position, hesitation behavior recognition model, each preset job type, personality trait analysis model, coefficient feature tuning values ​​corresponding to each personality trait scoring range, target students' liking coefficient for each preset job type, preset job type corresponding to each simulated test scenario, and adaptation pyramid of each preset job type.

[0021] It should also be noted that the various simulated test scenarios include: product design scenario, interface development scenario, code writing scenario, system testing scenario, operation and maintenance deployment scenario, user operation scenario, and project management scenario, and in all simulated test scenarios, target students are allowed to query questions through a browser.

[0022] In one specific embodiment, the task completion time and task completion quality score of the target student in each simulated test scenario are obtained by means of the target student's test terminal.

[0023] The system acquires the target student's typing data, task operation path records, and eye movement data at each monitoring time point during the monitoring period in each simulated test scenario. It also acquires the target student's personality dimension score data and uses the target student's personality dimension score data, typing data during the monitoring period in each simulated test scenario, and eye movement data at each monitoring time point as the target student's latent feature data.

[0024] It should be noted that the typing data includes: each deleted text, the duration of each typing pause, and the typing rhythm fluctuation value; the task operation path record includes: the time points of each operation interface transition and operation interface behavior; and the eye tracking data includes: fixation coordinates.

[0025] It should also be noted that the typing rhythm fluctuation value is a comparison between the target student's typing speed and their own average typing speed, excluding the time spent pausing while typing.

[0026] In one specific embodiment, the typing data, task operation path records, and eye-tracking data of the target student during the monitoring period in each simulated test scenario are obtained, and the score data of each personality dimension of the target student are obtained. The specific method is as follows: the typing data and task operation path records of the target student during the monitoring period in each simulated test scenario are obtained through the target student's testing terminal; the eye-tracking data of the target student at each monitoring time point in each simulated test scenario is obtained through the eye-tracking device worn by the target student; and the score data of each personality dimension of the target student is obtained through personality tests such as MBTI.

[0027] The student profile fitting module is used to perform multidimensional feature analysis based on the explicit and implicit feature data of the target student, thereby obtaining the explicit feature set and implicit feature set of the target student. Then, vector analysis is performed on the explicit feature set and implicit feature set of the target student to obtain the target feature profile of the target student.

[0028] In a specific embodiment of the present invention, the explicit feature set and the implicit feature set of the target student are obtained by means of: calculating the efficiency evaluation coefficient of the target student in each simulated test scenario based on the task completion time of the target student in each simulated test scenario and in combination with the task completion quality score of the target student in each simulated test scenario, and using it as the explicit feature set of the target student.

[0029] In one specific embodiment, the efficiency evaluation coefficient of the target student in each simulated test scenario is calculated as follows: The average task completion time A and average task completion quality score B are obtained from the local database, and the coefficients are calculated based on the target student's task completion time in each simulated test scenario. The assessment also incorporates the quality of student performance in various simulated test scenarios. , where x represents the number of each simulated test scenario, Let y be a positive integer greater than 2. Calculate the efficiency evaluation coefficient of the target student in each simulated test scenario. , where e is the natural constant.

[0030] It should be noted that the average task completion time and average task completion quality score were obtained by researchers based on the test results of a large number of testers.

[0031] Based on the task operation path records of the target students in each simulated test scenario, analyze the proficiency evaluation coefficient of the target students in each simulated test scenario.

[0032] Based on the eye-tracking data of the target students at each monitoring time point in each simulated test scenario, we analyzed the attention concentration coefficient of the target students in each simulated test scenario.

[0033] Based on the typing data of the target students during the monitoring period in each simulated test scenario, the operational hesitation coefficient of the target students in each simulated test scenario was analyzed.

[0034] The proficiency evaluation coefficient, attention concentration coefficient, and operational hesitation coefficient of the target students in various simulated test scenarios are used as the set of implicit characteristics of the target students.

[0035] In a specific embodiment of the present invention, the proficiency evaluation coefficient of the target student in each simulated test scenario is analyzed. The specific method is as follows: based on the task operation path record of the target student in each simulated test scenario, the time points of each operation interface transition of the target student in each simulated test scenario are extracted, and the operation interface behavior of the target student after each operation interface transition time point in each simulated test scenario is extracted.

[0036] It should be noted that the interface switching time point is the time point when the target student switches from the test interface to a browser interface.

[0037] The interface behavior analysis model is obtained from the local database. Based on the target student's interface behavior after each interface transition time point in each simulated test scenario, the purpose of the target student's transition at each interface transition time point in each simulated test scenario is obtained. Combined with the target student's interface transition time points in each simulated test scenario, the proficiency evaluation coefficient of the target student in each simulated test scenario is calculated.

[0038] It should be noted that the interface behavior analysis model is obtained through machine learning and is trained by researchers based on a large amount of interface operation data.

[0039] In one specific embodiment, the purpose of the target student's interface switching at each time point in each simulated test scenario is obtained. The specific method is as follows: The interface behavior analysis model observes the target student's specific behavior after switching the interface, such as entering search keywords in the browser, executing commands in the terminal, and pasting code in the editor. Using the behavior pattern recognition ability trained by machine learning, it automatically determines the true intention of the interface switching, such as: querying help, testing and verification, obtaining resources, invalid roaming, etc.

[0040] In one specific embodiment, the proficiency evaluation coefficient of the target student in each simulated test scenario is calculated as follows: The proficiency evaluation coefficient tuning values ​​corresponding to each conversion purpose are obtained from the local database. Based on the conversion purpose of the target student at each operation interface conversion time point in each simulated test scenario, the proficiency evaluation coefficient tuning values ​​of the target student at each operation interface conversion time point in each simulated test scenario are obtained. These values ​​are then summed to obtain the total proficiency evaluation coefficient tuning value of the target student in each simulated test scenario. Based on the operation interface conversion time point of the target student in each simulated test scenario, the number of interface switching times of the target student in each simulated test scenario is counted. The total proficiency evaluation coefficient tuning value and the number of interface switching times of the target student in each simulated test scenario are added together to obtain the proficiency evaluation coefficient of the target student in each simulated test scenario.

[0041] In a specific embodiment of the present invention, the attention concentration coefficient of the target student in each simulated test scenario is analyzed. The specific method is as follows: based on the eye movement data of the target student at each monitoring time point in each simulated test scenario, the gaze coordinates of the target student at each monitoring time point in each simulated test scenario are extracted.

[0042] The screen display analysis model and the screen images of the target student at each monitoring time point in each simulated test scenario are obtained from the local database. The screen images of the target student at each monitoring time point in each simulated test scenario are put into the screen display analysis model to obtain the heat map of the area distribution of the screen images of the target student at each monitoring time point in each simulated test scenario.

[0043] It should be noted that the screen display analysis model is obtained through machine learning and trained by researchers using a large number of screen images. The screen display analysis model intelligently analyzes screenshots taken during student tests and uses computer vision technology to automatically identify key information areas in the screen, such as code editing areas, error message boxes, task requirement text, and output areas of running results, and generates heat maps that mark the location and importance of these key areas.

[0044] Based on the gaze coordinates of the target students at each monitoring time point in each simulated test scenario, and combined with the heat map of the screen area distribution of the target students at each monitoring time point in each simulated test scenario, the attention concentration level of the target students at each monitoring time point in each simulated test scenario is evaluated, and the attention concentration coefficient of the target students in each simulated test scenario is calculated accordingly.

[0045] In one specific embodiment, the attention concentration level of the target student at each monitoring time point in each simulated test scenario is evaluated, and the attention concentration coefficient of the target student in each simulated test scenario is calculated accordingly. The specific method is as follows: based on the combined screen area distribution heatmap of the target student at each monitoring time point in each simulated test scenario, the key distribution level of each screen area of ​​the target student at each monitoring time point in each simulated test scenario is obtained; based on the gaze coordinates of the target student at each monitoring time point in each simulated test scenario, the key distribution level of the area gazed at by the target student at each monitoring time point in each simulated test scenario is determined, and this key distribution level is used as the attention concentration level of the target student at each monitoring time point in each simulated test scenario. Where n represents the number of each monitoring time point. Let m be a positive integer greater than 2. Calculate the attention concentration coefficient of the target student in each simulated test scenario. , where m is the number of monitoring time points.

[0046] In a specific embodiment of the present invention, the operational hesitation coefficient of the target student in each simulated test scenario is analyzed. The specific method is as follows: based on the typing data of the target student during the monitoring period in each simulated test scenario, the target student's text deletion, typing pause duration and typing rhythm fluctuation value in each simulated test scenario are extracted.

[0047] Based on the pause duration of the target student in each typing session in each simulated test scenario, the pause fluctuation coefficient of the target student in each simulated test scenario is calculated.

[0048] In one specific embodiment, the pause fluctuation coefficient of the target student in each simulated test scenario is calculated. The specific method is as follows: based on the pause duration of each typing session of the target student in each simulated test scenario. Where i represents the number of each typing pause, j is a positive integer greater than 2, and is combined with the task completion time of the target student in each simulated test scenario. Calculate the pause fluctuation coefficient of the target student in each simulated test scenario. .

[0049] Based on the text deleted by the target student in each simulated test scenario, the character length of the text deleted by the target student in each simulated test scenario is counted, thereby calculating the average length of the text deleted by the target student in each simulated test scenario.

[0050] Retrieve the final draft text corresponding to each deleted text position in the final draft of each simulated test scenario from the local database, and mark it as the deleted and replaced text of each simulated test scenario. Compare the semantic similarity between the deleted text and the deleted and replaced text of each simulated test scenario, and calculate the semantic deviation coefficient of the deleted text of the target student in each simulated test scenario.

[0051] In one specific embodiment, the semantic deviation coefficient of deleted text for the target student in each simulated test scenario is calculated. The specific method is as follows: the existing semantic deviation analysis technology is relatively mature, and the semantic deviation coefficient of deleted text for the target student in each simulated test scenario can be obtained through the existing semantic deviation analysis technology.

[0052] The hesitation behavior recognition model is obtained from the local database. The pause fluctuation coefficient, typing rhythm fluctuation value, average length of deleted text, and semantic deviation coefficient of deleted text of the target student in each simulated test scenario are input into the hesitation behavior recognition model to obtain the hesitation behavior feature score of the target student in each simulated test scenario, and it is used as the operational hesitation coefficient of the target student in each simulated test scenario.

[0053] It should be noted that the hesitation behavior recognition model, obtained through machine learning, comprehensively analyzes the multidimensional hesitation characteristics of students during typing: the pause fluctuation coefficient reflects the frequency and irregularity of thought interruption; the typing rhythm fluctuation value reflects the instability of input speed; the average length of deleted text indicates the extent of regression when rejecting one's own ideas; and the semantic deviation coefficient of deleted text measures the degree of semantic difference between the deleted content and the final draft. A large deviation indicates repeated changes in thinking. The model uses these four features as input vectors and calculates a comprehensive hesitation behavior feature score through trained neural network weights. This score is the operational hesitation coefficient, used to assess students' implicit abilities at the decision-making and execution levels.

[0054] The job feature and student profile matching module is used to obtain each preset job type from the local database, and calculate the matching recommendation index between the target student and each preset job type based on the target student's target feature profile and a hybrid recommendation analysis method.

[0055] In a specific embodiment of the present invention, the target feature profile of the target student is obtained by means of: obtaining the personality feature analysis model and the parameter tuning values ​​of each coefficient feature corresponding to each personality feature score interval from the local database; analyzing the personality feature score of the target student based on the score data of each personality dimension of the target student; mapping the parameter tuning values ​​of the latent feature of the target student based on the parameter tuning values ​​corresponding to each personality feature score interval; and adjusting the parameters accordingly to obtain the set of adjusted latent features of the target student.

[0056] It should be noted that existing personality trait analysis models are relatively mature, and personality trait scores for target students can be obtained based on these models.

[0057] It should also be noted that the parameter adjustment values ​​for each coefficient corresponding to the score interval of each personality trait refer to the pre-set characteristic coefficient correction parameters based on the influence of different personality types on work performance. For example, students with high extroversion scores may have a high operational hesitation coefficient in operational scenarios, but this may be due to their ability to think divergently rather than genuine hesitation. Therefore, the operational hesitation parameter adjustment value for this interval is set to 0.8 to reduce the negative impact of the hesitation coefficient. On the other hand, students with high conscientiousness scores have a more valuable higher attention concentration coefficient in testing scenarios. The attention concentration parameter adjustment value for this interval is set to 1.2 to amplify the attention advantage. This value is adjusted by researchers according to the actual situation.

[0058] In one specific embodiment, the parameter tuning obtains the set of latent features of the target student. The specific method is as follows: multiply the proficiency evaluation coefficient, attention concentration coefficient and operational hesitation coefficient of the target student in each simulated test scenario by the latent feature parameter tuning value to calculate the adjustment proficiency evaluation coefficient, adjustment attention concentration coefficient and adjustment operational hesitation coefficient of the target student in each simulated test scenario, and use them as the set of latent features of the target student.

[0059] Based on the explicit feature set and adjusted implicit feature set of the target student, analyze the three-dimensional feature cone of the target student and use it as the target feature profile of the target student.

[0060] In a specific embodiment of the present invention, the three-dimensional feature cone of the target student is analyzed. The specific method is as follows: based on the target student's set of adjustment latent features, the adjustment proficiency evaluation coefficient, adjustment attention concentration coefficient, and adjustment operation hesitation coefficient of the target student in each simulated test scenario are extracted, and the above three coefficients are combined to form the special bottom triangle of the target student in each simulated test scenario. The center point of the special bottom triangle of the target student in each simulated test scenario is extracted.

[0061] It should be noted that, referring to Figure 2 As shown, Figure 2 The angles formed between each pair of 1, 2, and 3 are 120°. The lengths of 1, 2, and 3 are adjusted based on the actual data of the adjustment proficiency evaluation coefficient, adjustment attention concentration coefficient, and adjustment operation hesitation coefficient. 4 is the starting point of 1, 2, and 3, and the endpoints of 1, 2, and 3 are the three vertices of the special bottom triangle.

[0062] Based on the explicit feature set of the target student, the efficiency evaluation coefficient of the target student in each simulated test scenario is extracted. The extension is perpendicularly outward from the center point of the special bottom triangle of the target student in each simulated test scenario, and the extension length is the efficiency evaluation coefficient of the corresponding simulated test scenario. This determines the fourth vertex of the target student in each simulated test scenario, and connects it with the three vertices of the special bottom triangle of the target student in each simulated test scenario. This results in the special triangular pyramid of the target student in each simulated test scenario, which is used as the three-dimensional feature pyramid of the target student.

[0063] The job recommendation processing module is used to generate a job recommendation list for the target student based on the matching recommendation index between the target student and each preset job type, and send it to the target student for visual recommendation.

[0064] In a specific embodiment of the present invention, the matching recommendation index between the target student and each preset job type is calculated. The specific method is as follows: based on the target student's target feature profile, the profile adaptation coefficient between the target student and each preset job type is calculated.

[0065] The system retrieves the target students' preference coefficients for each preset job type from the local database, and calculates the matching recommendation index between the target students and each preset job type based on the profile adaptation coefficients of the target students and each preset job type.

[0066] It should be noted that the target students' preference for each preset job type was obtained through a questionnaire.

[0067] In one specific embodiment, the matching recommendation index between the target student and each preset job type is calculated by multiplying the target student's liking coefficient for each preset job type by the profile adaptation coefficient.

[0068] In a specific embodiment of the present invention, the profile adaptation coefficient between the target student and each preset job type is calculated. The specific calculation method is as follows: obtain the preset job type corresponding to each simulated test scenario and the adaptation pyramid of each preset job type from the local database, map the adaptation pyramid of the target student in each simulated test scenario, compare the similarity between the special pyramid of the target student in each simulated test scenario and the adaptation pyramid, thereby obtaining the pyramid similarity of the target student in each simulated test scenario, and obtain the pyramid similarity between the target student and each preset job type according to the preset job type corresponding to each simulated test scenario, and use it as the profile adaptation coefficient between the target student and each preset job type.

[0069] It should be noted that the preset job types corresponding to each simulated test scenario refer to the mapping relationship between scenarios and jobs pre-set by the system. For example, the product design scenario corresponds to the product manager job, the interface development scenario corresponds to the UI designer job, and the code writing scenario corresponds to the front-end / back-end development job. The adaptation pyramid of each preset job type is a standard profile pyramid constructed by statistically analyzing historical data of successfully employed students. The construction process of the standard profile pyramid is consistent with the method of analyzing the three-dimensional feature pyramid of the target students.

[0070] In one specific embodiment, the pyramid similarity of the target student in each simulated test scenario is obtained. The specific method is as follows: the existing pyramid similarity comparison technology is relatively mature. The pyramid similarity of the target student in each simulated test scenario can be obtained through the existing pyramid similarity comparison technology.

[0071] In one specific embodiment, the method for generating a job recommendation list for target students is as follows: sort the preset job types according to the matching recommendation index from largest to smallest, thereby obtaining the preset job types after sorting the target students, and using them as the job recommendation list for target students.

[0072] This invention constructs a multi-level, three-dimensional student characteristic assessment system, significantly improving the objectivity, accuracy, and dimensionality of student ability characteristic expression. The system can comprehensively capture students' explicit work abilities and implicit professional qualities, forming a more realistic and accurate ability profile. This avoids the subjective bias and information distortion problems of traditional self-assessment questionnaires, providing a scientific data foundation for subsequent job matching. This invention can deeply mine behavioral characteristics and cognitive patterns during task execution, significantly improving the accuracy of ability assessment and the suitability of job matching. Simultaneously, this invention implements a personalized job assessment strategy and, based on a geometric modeling method of three-dimensional feature cones, transforms abstract multi-dimensional ability characteristics into a visualized and quantifiable spatial profile, achieving scientific and precise person-job matching. Furthermore, it combines students' job preference coefficients for comprehensive recommendations, balancing objective ability suitability and subjective willingness guidance, improving the acceptance of recommendation results and employment success rate.

[0073] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0074] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A recommendation system based on machine learning and matching multidimensional features with job profiles, characterized in that, include: The multidimensional feature acquisition module is used to conduct simulated assessments of target students in scientific research enterprises, thereby acquiring explicit and implicit feature data of the target students. The student profile fitting module is used to perform multi-dimensional feature analysis based on the explicit and implicit feature data of the target student, thereby obtaining the explicit feature set and implicit feature set of the target student. Then, vector analysis is performed on the explicit feature set and implicit feature set of the target student to obtain the target feature profile of the target student. The specific method for obtaining the set of explicit features and the set of implicit features of the target student is as follows: Based on the task completion time of the target students in each simulated test scenario, and combined with the task completion quality score of the target students in each simulated test scenario, the efficiency evaluation coefficient of the target students in each simulated test scenario is calculated, and it is used as the explicit feature set of the target students. Based on the task operation path records of the target students in each simulated test scenario, analyze the proficiency evaluation coefficient of the target students in each simulated test scenario; Based on the eye-tracking data of the target students at each monitoring time point in each simulated test scenario, the attention concentration coefficient of the target students in each simulated test scenario was analyzed. Based on the typing data of the target students during the monitoring period in each simulated test scenario, the operational hesitation coefficient of the target students in each simulated test scenario was analyzed. The proficiency evaluation coefficient, attention concentration coefficient, and operational hesitation coefficient of the target students in various simulated test scenarios are used as the set of implicit characteristics of the target students. The specific method for analyzing the operational hesitation coefficient of the target students in various simulated test scenarios is as follows: Based on the typing data of the target students during the monitoring period of each simulated test scenario, extract the text deletion, typing pause duration and typing rhythm fluctuation value of the target students in each simulated test scenario. Based on the pause duration of the target student in each typing session in each simulated test scenario, calculate the pause fluctuation coefficient of the target student in each simulated test scenario. Based on the text deleted by the target student in each simulated test scenario, the character length of the text deleted by the target student in each simulated test scenario is counted, and the average length of the text deleted by the target student in each simulated test scenario is calculated. Retrieve the final text corresponding to each deleted text position in the final draft of the target student in each simulated test scenario from the local database, and mark it as the deleted and replaced text of the target student in each simulated test scenario. Compare the semantic similarity between the deleted text of the target student in each simulated test scenario and the deleted and replaced text of each simulated test scenario, and calculate the semantic deviation coefficient of the deleted text of the target student in each simulated test scenario. The hesitation behavior recognition model is obtained from the local database. The pause fluctuation coefficient, typing rhythm fluctuation value, average length of deleted text and semantic deviation coefficient of deleted text of the target student in each simulated test scenario are input into the hesitation behavior recognition model to obtain the hesitation behavior feature score of the target student in each simulated test scenario, and it is used as the operational hesitation coefficient of the target student in each simulated test scenario. The specific method for obtaining the target feature profile of the target student is as follows: The system retrieves the personality trait analysis model and the parameter tuning values ​​of each coefficient corresponding to each personality trait score interval from the local database. Based on the score data of each personality dimension of the target student, it analyzes and obtains the personality trait score of the target student. Based on the parameter tuning values ​​of each personality trait score interval, it maps the parameter tuning values ​​of the target student's latent traits and obtains the adjusted latent trait set of the target student based on these parameter tuning values. Based on the explicit feature set and adjusted latent feature set of the target student, analyze the three-dimensional feature cone of the target student and use it as the target feature profile of the target student; The specific method for analyzing the three-dimensional feature cone of the target student is as follows: Based on the target student's set of latent features, the target student's adjustment proficiency evaluation coefficient, adjustment attention concentration coefficient, and adjustment operation hesitation coefficient in each simulated test scenario are extracted. The target student's special bottom triangle in each simulated test scenario is then combined based on the above three coefficients, and the center point of the target student's special bottom triangle in each simulated test scenario is extracted. Based on the explicit feature set of the target student, the efficiency evaluation coefficient of the target student in each simulated test scenario is extracted. The center point of the special bottom triangle of the target student in each simulated test scenario is extended vertically outward, and the extension length is the efficiency evaluation coefficient of the corresponding simulated test scenario. This determines the fourth vertex of the target student in each simulated test scenario. The fourth vertex is then connected to the three vertices of the special bottom triangle of the target student in each simulated test scenario to obtain the special triangular pyramid of the target student in each simulated test scenario. This pyramid is then used as the three-dimensional feature pyramid of the target student. The job feature and student profile matching module is used to obtain each preset job type from the local database, and calculate the matching recommendation index between the target student and each preset job type based on the target student's target feature profile and a hybrid recommendation analysis method. The job recommendation processing module is used to generate a job recommendation list for the target student based on the matching recommendation index between the target student and each preset job type, and send it to the target student for visual recommendation.

2. The recommendation system based on machine learning and matching multidimensional features with job profiles according to claim 1, characterized in that, The specific method for obtaining the explicit and implicit feature data of the target student is as follows: The simulated test scenarios of scientific research enterprises are obtained from the local database. The task completion time and task completion quality score of the target students in each simulated test scenario are obtained and used as the explicit characteristic data of the target students. The system acquires the target student's typing data, task operation path records, and eye movement data at each monitoring time point during the monitoring period in each simulated test scenario. It also acquires the target student's personality dimension score data and uses the target student's personality dimension score data, typing data during the monitoring period in each simulated test scenario, and eye movement data at each monitoring time point as the target student's latent feature data.

3. The recommendation system based on machine learning and matching multidimensional features with job profiles according to claim 1, characterized in that, The specific method for analyzing the proficiency evaluation coefficients of the target students in each simulated test scenario is as follows: Based on the task operation path records of the target students in each simulated test scenario, extract the time points of each operation interface transition of the target students in each simulated test scenario, and extract the operation interface behavior of the target students after each operation interface transition time point in each simulated test scenario. The interface behavior analysis model is obtained from the local database. Based on the target student's interface behavior after each interface transition time point in each simulated test scenario, the purpose of the target student's transition at each interface transition time point in each simulated test scenario is obtained. Combined with the target student's interface transition time points in each simulated test scenario, the proficiency evaluation coefficient of the target student in each simulated test scenario is calculated.

4. The recommendation system based on machine learning and matching multidimensional features with job profiles according to claim 3, characterized in that, The specific method for analyzing the attention concentration coefficient of the target students in each simulated test scenario is as follows: Based on the eye-tracking data of the target student at each monitoring time point in each simulated test scenario, the fixation coordinates of the target student at each monitoring time point in each simulated test scenario are extracted. The screen image analysis model and the screen images of the target student at each monitoring time point in each simulated test scenario are obtained from the local database. The screen images of the target student at each monitoring time point in each simulated test scenario are put into the screen image analysis model to obtain the heat map of the area distribution of the screen images of the target student at each monitoring time point in each simulated test scenario. Based on the gaze coordinates of the target students at each monitoring time point in each simulated test scenario, and combined with the heat map of the screen area distribution of the target students at each monitoring time point in each simulated test scenario, the attention concentration level of the target students at each monitoring time point in each simulated test scenario is evaluated, and the attention concentration coefficient of the target students in each simulated test scenario is calculated accordingly.

5. The recommendation system based on machine learning and matching multidimensional features with job profiles according to claim 1, characterized in that, The specific method for calculating the matching recommendation index between the target students and each preset job type is as follows: Based on the target student's profile, calculate the profile fit coefficient between the target student and each preset job type; The system retrieves the target students' preference coefficients for each preset job type from the local database, and calculates the matching recommendation index between the target students and each preset job type based on the profile adaptation coefficients of the target students and each preset job type.

6. The recommendation system based on machine learning and matching multidimensional features with job profiles according to claim 5, characterized in that, The specific calculation method for the adaptability coefficient between the target student profile and each preset job type is as follows: The system retrieves the preset job types and adaptive pyramids corresponding to each simulated test scenario from the local database. It then maps these to the adaptive pyramids of the target student in each simulated test scenario. The system compares the similarity between the target student's specific pyramids and the adaptive pyramids in each simulated test scenario to obtain the pyramid similarity of the target student in each simulated test scenario. Based on the preset job types corresponding to each simulated test scenario, the system obtains the pyramid similarity between the target student and each preset job type, and uses this as the profile adaptation coefficient between the target student and each preset job type.

Images