A double-teacher recommendation system based on big data analysis

Abstract

The application discloses a double-teacher recommendation system based on big data analysis, constructs a structured evidence pool through OCR analysis, digital signature verification and hash check, adopts a 2-of-3 gate valve mechanism to ensure the authenticity of data, carries out forced screening according to the education background and working years, calculates a bottleneck value through teaching experience, the number of training projects and the feedback score of students, divides the ability level, carries out double-key hard door screening based on the degree of professional matching and the practical years, removes dominated solutions in combination with a Pareto skyline, and determines a recommendation list through dictionary sequence sorting, adopts an analytic hierarchy process to construct a judgment matrix, calculates the weight and the comprehensive score, and supports two output strategies of default and weighting. The system effectively improves the accuracy, interpretability and reliability of teacher qualification authenticity verification and person-post matching, and is suitable for vocational education and enterprise training scenes.

Description

Technical Field

[0001] This invention relates to the field of big data analytics, and more specifically, to a dual-qualified teacher recommendation system based on big data analytics. Background Technology

[0002] Dual-qualified teachers refer to a teaching force that possesses both solid theoretical teaching skills and rich practical experience and expertise. Current systems and practices primarily focus on "teacher assessment" rather than "teacher matching." Teacher performance evaluation systems emphasize indicators such as teaching workload, research achievements, or classroom evaluation, but lack effective person-job matching tools, making it difficult to quickly identify the most suitable mentors for course offerings, industry-university collaborative projects, or innovative teaching projects. Compared to the goals of vocational education reform in the new era, problems remain, including insufficient numbers, a single source of qualified teachers, limited two-way flow between schools and enterprises, and a shortage of teachers with both theoretical teaching and practical guidance abilities. Therefore, the existing assessment system cannot meet the needs of the new era, and there is an urgent need to construct a recommendation system that translates evaluation results into practical application scenarios.

[0003] To address the above problems, this invention proposes a solution. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a dual-qualified teacher recommendation system based on big data analysis to address the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A dual-qualified teacher recommendation system based on big data analysis includes: an authenticity verification module, a teacher feature evaluation module, a matching and screening module, and a constraint control module, with signal connections between the modules; The authenticity verification module acquires multi-source evidence files from teachers, constructs a structured evidence pool through OCR parsing, digital signature verification, and hash verification, and then verifies the authenticity of the evidence. The teacher characteristic assessment module uses pre-set mandatory conditions to screen qualified teachers. Then, it collects three indicators: the instructor's teaching experience, the number of training projects, and the average feedback score from trainees. The minimum value among the three is taken as the instructor's bottleneck value to determine the level of their ability bottleneck. The matching and filtering module uses a dual-key hard-door screening based on professional relevance and years of practical experience, combined with Pareto skyline to retain non-dominated solutions, and obtains a candidate list for matching and filtering after sorting by lexicographical order. The constraint control module constructs a three-layer judgment matrix, generates weights and calculates the comprehensive score through the analytic hierarchy process, and provides two output strategies: default and weighted.

[0006] In a preferred embodiment, the authenticity verification module includes the following steps:

[0007] The authenticity verification adopts a 2-of-3 authenticity gate mechanism, which can determine the authenticity of the data by passing any two of the following: OCR text extraction verification, digital signature and timestamp verification, and document hash verification.

[0008] In a preferred embodiment, the teacher characteristic assessment module includes the following steps: The mandatory items for teacher characteristic assessment include educational background and years of work experience. The educational background requires a master's degree or above and the graduating institution must be registered with the Ministry of Education. The years of work experience must be at least a minimum. In the teacher characteristic assessment, the bottleneck value is the minimum value of teaching experience, number of training projects and trainee feedback scores after normalization. Based on this, teachers are divided into four competency levels: Level 1, Level 2, Level 3 and Level 4.

[0009] In a preferred embodiment, the matching and filtering module includes the following steps: In the dual-key hard-door screening of matching and filtering, the professional relevance is calculated by using the TF-IDF model to calculate the cosine similarity between course requirements and teacher background, and the practical years need to be verified by checking the project contract to confirm that the cumulative years meet the requirements; Pareto skyline matching filter constructs feature vectors based on years of teaching experience, student feedback ratings, and number of training programs, eliminating dominated and disadvantageous candidates. The matching and filtering process prioritizes comparing the average score of student feedback based on lexicographical order, followed by comparing the number of training programs and years of teaching experience.

[0010] In a preferred embodiment, the constraint control module includes the following steps:

[0011] In constraint control, reliability, coverage, and identifiability are calculated for the four major modules of layer A, the sub-modules of layer B, and the sub-items of layer C, and a judgment matrix is ​​generated. In constraint control, the consistency of a matrix is ​​judged by the consistency index CI and the random consistency ratio CR. < Normalized feature vectors are used as weights. The weighted strategy of constraint control is only enabled when all decision matrices meet the consistency requirement; otherwise, it automatically reverts to the default strategy and outputs the decision results of the first three modules.

[0012] The technical effects and advantages of the dual-qualified teacher recommendation system based on big data analysis proposed in this invention are as follows: By introducing OCR and hash signature technologies and establishing a structured evidence pool, the problems of forgery and omissions in traditional record management are avoided, and the authenticity of teachers' qualifications and achievements can be quickly verified.

[0013] Aligning and matching supply and demand indicators with isomorphic characteristics, the teacher evaluation indicators are projected onto the same detailed item set as the curriculum and job requirements, and a fixed dictionary and threshold are used to ensure consistency of dimensions; this enables the evaluation results to be directly used for matching, improving the scientific nature of the person-job matching.

[0014] Each decision-making stage outputs a Boolean state, indicator hit status, and corresponding evidence chain, making it easier for teaching supervisors or companies to understand the reasons for the recommendation and to audit it; all parameters and rules are public and can be manually recalculated, meeting the needs of patent disclosure and industry supervision.

[0015] This paper proposes an implementable recommendation framework by combining k-of-n rules, bottleneck method, Pareto skyline, lexicographical order, and evaluation matching of dual-qualified teachers. It maintains the structure of the traditional evaluation system and avoids the drawbacks of weighted compensation. It provides model backtracking and manual review mechanisms, creating a new paradigm for recommendation in the education industry.

[0016] It overcomes the shortcomings of existing technologies in teacher matching, and ensures that the matching results are scientific, fair, and reproducible through unweighted and interpretable decision-making logic, thus possessing significant social value and potential for widespread application. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of a dual-qualified teacher recommendation system based on big data analysis according to the present invention.

[0018] Figure 2 This is a schematic diagram of a dual-qualified teacher recommendation system module based on big data analysis according to the present invention. Detailed Implementation

[0019] 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.

[0020] Example 1: Please refer to Figures 1-2 As shown, this invention discloses a dual-qualified teacher recommendation system based on big data analysis, including: an authenticity verification module, a teacher feature evaluation module, a matching and screening module, and a constraint control module, with signal connections between the modules; Authenticity Verification Module: Obtains multi-source evidence files from teachers, constructs a structured evidence pool through OCR parsing, digital signature verification, and hash verification, and performs authenticity verification. Teacher Characteristic Assessment Module: Based on preset mandatory conditions, qualified teachers are screened and reviewed. Then, the minimum value of the three indicators (teacher experience, number of training projects, and average feedback score of trainees) is taken as the bottleneck value of the teacher to determine the level of the teacher's ability bottleneck. Matching and filtering module: Based on professional relevance and years of practical experience, a dual-key hard-door screening is performed. The Pareto skyline is used to retain non-dominated solutions. The candidate list is obtained by sorting by lexicographical order and then matched and filtered. Constraint control module: Constructs a three-layer judgment matrix, generates weights and calculates the comprehensive score through the analytic hierarchy process, and provides two output strategies: default and weighted.

[0021] In the authenticity verification module, multi-source evidence files from teachers are obtained. A structured evidence pool is constructed through OCR parsing, digital signature verification, and hash checking for authenticity verification. Specific content includes: Acquire multi-source data including various evidentiary documents such as teachers' academic degree certificates, professional title certificates, professional qualification certificates, work experience certificates, teaching materials, scientific research achievements, technical service contracts, and student feedback evaluations. Use the TesseractOCR engine to convert images or scanned documents into readable text. Record metadata such as source, uploader, timestamp, page number, and file hash value for each piece of evidence to build a structured evidence database composed of multi-source evidentiary documents. The "2-of-3 authenticity valve" mechanism is adopted: data is deemed authentic and valid if any three of the following four verifications are passed: OCR text extraction verification: The TesseractOCR engine is used to perform text recognition on the certificate image. The overall recognition confidence level is required to be greater than or equal to the confidence threshold. Based on this, it is determined whether the extracted text information of the certificate is reliable. Digital signature and timestamp verification: For electronic certificate files, verify the attached digital signature and trusted timestamp. The signature uses the RSA2048 algorithm and is issued by a nationally certified timestamp service center. Use a public key to verify whether the digital signature is valid and confirm that the timestamp is within the validity period and has not been tampered with. Document hash verification: Calculate the SHA-256 hash value for each uploaded multi-source evidence file and compare it with the original hash recorded at the time of upload. If the hash values ​​match, it proves that the file has not been modified since the upload and has maintained its integrity and authenticity. If less than two of the above three checks are passed, the lecturer's information will be considered to have a risk of authenticity and will be automatically placed into the manual review queue and transferred to the manual review process. The reviewers will further verify the authenticity of the information and record the conclusions. Then, based on the manual conclusions, they will decide whether to proceed to the subsequent evaluation process. After the above processing, this module outputs a list of available evidence and a list of evidence pending review, and provides a "qualified / unqualified" status for each teacher on each mandatory evidence item.

[0022] In the teacher characteristic assessment module, qualified teachers are screened based on pre-set mandatory conditions. Then, by collecting data on three indicators—teaching experience, number of training projects, and average student feedback score—the minimum value among these three is taken as the instructor's bottleneck value to determine their competency level. Specific content includes: Teachers are evaluated according to the indicator system, which is divided into four modules and sub-items: background matching, resource utilization, teaching process, and results and benefits. Methods such as keyword identification, frequency statistics, and gradient scoring are used to assign values ​​to each sub-item.

[0023] For example, a doctoral degree is worth 100 points, a master's degree is worth 80 points, and a bachelor's degree is worth 60 points; 10 years of work experience is worth 100 points; using virtual simulation technology in teaching is worth 20 points, and using multimedia teaching is worth 15 points, etc. The scores are normalized to the range of 0 to 100. After the information is verified for authenticity, the key professional characteristics of the lecturers are evaluated and screened, and a mandatory system is established, which includes two parts: mandatory qualification review and determination of ability bottleneck level. The mandatory qualification review retains the following two essential conditions: educational background and years of work experience. Both must be met. If either mandatory condition is not met, the lecturer is deemed unqualified and immediately eliminated, not proceeding to the next evaluation stage. Further competency assessment will only be conducted if both conditions are met. Educational Background: Lecturers must have a master's degree or above, and their alma mater must be listed in the list of universities registered with the Ministry of Education. The information on the academic certificates provided by the lecturers will be compared with the authoritative database of the Ministry of Education to verify whether the academic level and school qualifications meet the requirements. Years of work experience: The lecturer's cumulative years of work experience shall not be less than the minimum years. The years of work experience shall be calculated based on the social security payment records or scanned copies of the labor contract uploaded by the lecturer to determine whether the minimum years of work experience have been reached. If any of the above mandatory requirements are not met, the lecturer will be deemed not to meet the basic qualification requirements and will be directly eliminated, and will not proceed to the subsequent evaluation process; only if both conditions are met will the lecturer undergo further competency evaluation.

[0024] After the mandatory qualification review is passed, the bottleneck level is determined. For instructors who pass the mandatory qualification review, their teaching-related indicators are further evaluated to determine their bottleneck level, which includes Level 1, Level 2, Level 3, and Level 4. Specifically, three indicators are collected: the instructor's teaching experience, the number of training projects, and the average feedback score from trainees. These three indicators are converted to a uniform score range of 0-100 using a linear normalization method, and the minimum value among the three is taken as the instructor's bottleneck value P, representing the degree of their skill deficiency. Based on the bottleneck value P, the instructor's ability is classified according to predetermined rules as follows: Level 1: Bottleneck value P , A score of 85 can be set, indicating that the instructor's abilities in all aspects are quite outstanding. Level 2: point, A score of 70 indicates that the instructor's overall ability is good, but there is a slight deficiency in one aspect. Level 3: point, A score of 60 indicates that the instructor has significant weaknesses and their overall ability is average. Level 4: A score of 0 indicates that the instructor's score is significantly low in at least one key indicator, and this score is not recommended.

[0025] If an instructor is rated as Level 4, they are usually not included in the candidate recommendation list; instructors at Levels 1 to 3 will proceed to the next matching and screening process, which ensures that the evaluation results objectively reflect the weakest link in the instructor's ability, making it easier to conduct more targeted screening in the matching stage. This step outputs each teacher's "qualified / unqualified" Boolean status, grade classification (Level 1, Level 2, Level 3, and Level 4), and whether or not they have been promoted, using gate valves, minimum values, and counting rules.

[0026] In the matching and filtering module, a dual-key hard-door screening is performed based on professional relevance and years of practical experience. This is combined with Pareto skyline preservation of non-dominated solutions, and the candidate list is obtained by lexicographical sorting for matching and filtering. Specific details include: The most suitable candidate for specific training needs is selected from the instructors who have passed the preliminary screening. A mechanism combining a "double-key hard-door" filtering system and multiple priority rankings is employed, specifically including: Dual-key hard-door screening: Candidate lecturers must meet both of the following hard criteria to pass the screening: Professional Relevance: The system uses a TF-IDF model to compare the specific course requirements description with the lecturer's background information text, calculating the cosine similarity score between the two to quantify the degree of matching in professional direction. If the similarity score is not lower than the similarity threshold, the lecturer's professional field is considered to be highly consistent with the course requirements; Years of practical experience: Instructors must have at least the minimum number of years of practical experience in actual projects. This will be confirmed by verifying the project contracts provided by the instructors to determine whether their accumulated years of practical experience have reached or exceeded the minimum number of years of practical experience. Only when both of the above conditions are met can the candidate lecturer pass the hard screening and enter the next step of the Skyline retention screening; Skyline Retention Screening: For candidate lecturers who pass the dual-key hard-gate screening, feature vectors are constructed based on three dimensions: "years of teaching experience, student feedback ratings, and number of training projects," and the Pareto optimality principle is used for screening. Specifically, all candidate vectors are compared, and any inferior candidate that is equal to or comprehensively surpassed by other candidates in all three indicators is eliminated; Distance: If the feature vectors constructed by candidate a in all three dimensions are not inferior to those of candidate b and are superior to b in at least one dimension, then a dominates b; the dominated ones are eliminated, thus preserving the skyline set. This process ensures that the remaining teachers have an advantage in at least one dimension and do not lag behind in other dimensions. After this "skyline" screening, the remaining lecturers are all non-dominated solutions that have unique advantages in certain indicators and have not been completely surpassed by others, thus ensuring that the candidates who enter the final ranking each have their own advantages and differences. Lexicographical Priority Sorting: For the candidate lecturer set obtained through the Skyline screening, the system sorts them according to a predetermined priority rule to select the best N lecturers to recommend to the user. The sorting uses a lexicographical cascade comparison method: first, sorting is done from highest to lowest based on the average student feedback score S; if feedback scores are tied, the number of training programs P is compared, with those having more programs ranking higher; if both feedback scores and the number of programs are the same, then the years of teaching experience T are compared, with those having more experience taking priority. After the above multiple priority sorting, the system selects the top N lecturers as the final recommendation list.

[0027] This step outputs a set of skylines selected by dual-key filtering and Pareto skylines, and obtains the Top-K ranking and key hit and key indicator explanations for each candidate in lexicographical order.

[0028] In the constraint control module, a three-layer judgment matrix is ​​constructed. Weights are generated and a comprehensive score is calculated using the analytic hierarchy process (AHP). Two output strategies are provided: default and weighted. Specific details include: The observables are defined by each layer of nodes, including the four modules of layer A, each sub-module of layer B, and each sub-item of layer C. A judgment matrix is ​​automatically generated. For any node X, three types of meta-indicators are defined within an evaluation batch, all of which are normalized to... : Using a data-driven Analytic Hierarchy Process (AHP) weight generation and total score summarization mechanism, the system automatically calculates pairwise comparison matrices. When the consistency check passes, the comprehensive score can be output. If the consistency check fails or this step is not enabled, the results of the first three modules are retained. Each node X includes layers A, B, and C. Layer A consists of three major modules: background matching, resource utilization, teaching process, and outcome benefits, which correspond to the four major modules in the teacher characteristic assessment module. Layer B consists of sub-modules, and Layer C consists of sub-items. Node X is a placeholder that refers to any node of the evaluated object in this evaluation hierarchy, and can fall in any of the three levels: Reliability The minimum of the mean confidence score and the evidence binding rate is taken. The mean confidence score is the average contextual confidence score of all valid hits for that node, and the evidence binding rate is the percentage of hits that can be bound to evidence IDs. If there are no valid hits for that node in this batch, then... To guarantee a minimum return; The evidence refers to the various criteria in the first three steps, such as the result of comparing the confidence level with the confidence threshold, whether the SHA-256 hash values ​​are consistent, and whether the accumulated professional experience is not less than the minimum number of years. Coverage : This reflects the percentage of lecturers who meet the condition that evidence ≥ Th: Distinctiveness The interquartile range (IQR) of the node's score is used to measure discriminative ability. ;in, This is the score of the node for lecturer u; For two nodes at the same level Calculate the set of ratios for the above indicators respectively. Take the median as the composite ratio. To prevent the divisor from being zero, 0.05 is used instead of 0 when the denominator is 0; Then Mapping to the Saaty1–9 scale: like <1 item =1; like If ≥1, then the intervals [1,1.25)→1, [1.25,1.5)→2, [1.5,1.75)→3, [1.75,2.0)→4, [2.0,2.5)→5, [2.5,3.0)→6, [3.0,3.5)→7, [3.5,4.0)→8, [4.0,∞)→9 are rounded down to obtain the result. This automatically generates judgment matrices of different sizes for layers A (4×4), B (3×3 each), and C (each with its own size). For each judgment matrix, calculate the largest eigenvalue. And the corresponding feature vectors, and normalize the feature vectors into weight vectors. Through consistency indicators: , Where n is the matrix order, For the random consistency factor, please refer to the random consistency factor table. < If the matrix consistency is considered satisfactory, its normalized eigenvectors can be used as the relative weights of the nodes in that layer; if CR ≥ The system prompts that additional evidence or adjustments to the judgment criteria are required, and then returns to the authenticity verification module. Let the weights of each module in layer A be... The weight of each submodule in layer B under its corresponding module in layer A is: The weight of each sub-item in layer C under its corresponding sub-module in layer B is: Then the weights of each sub-module and sub-item to the overall goal are as follows: , ;in Belonging to , Belonging to After the calculation is complete, a consistency check is performed again on the global weight vector. If the consistency is satisfied... < Then the weights are solidified; remember The final comprehensive score is calculated as follows: (The score is the value of the kth sub-item in layer C obtained from the first three steps.) ; Based on the above results, two output strategies are provided: Default strategy: Ignore the constraint control module and directly use the decision results of the first three modules as the final recommendation; Weighted strategy: When the user enables this module and all judgment matrices are satisfied. < At that time, output the weighted composite score. If any matrix does not meet the consistency requirement, it will automatically fall back to the default strategy.

[0029] The above formulas are all dimensionless calculations. 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 in the formulas are set by those skilled in the art according to the actual situation.

[0030] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other 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.

[0031] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0032] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0033] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0034] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dual-qualified teacher recommendation system based on big data analysis, characterized in that, Signal connections between modules; The authenticity verification module acquires multi-source evidence files from teachers, constructs a structured evidence pool through OCR parsing, digital signature verification, and hash verification, and then verifies the authenticity of the evidence. The teacher characteristic assessment module uses pre-set mandatory conditions to screen qualified teachers. Then, it collects three indicators: the instructor's teaching experience, the number of training projects, and the average feedback score from trainees. The minimum value among the three is taken as the instructor's bottleneck value to determine the level of their ability bottleneck. The matching and filtering module uses a dual-key hard-door screening based on professional relevance and years of practical experience, combined with Pareto skyline to retain non-dominated solutions, and obtains a candidate list for matching and filtering after sorting by lexicographical order. The constraint control module constructs a three-layer judgment matrix, generates weights and calculates the comprehensive score through the analytic hierarchy process, and provides two output strategies: default and weighted.

2. The dual-qualified teacher recommendation system based on big data analysis according to claim 1, characterized in that, The authenticity verification adopts a 2-of-3 authenticity gate mechanism, which can determine the authenticity of the data by passing any two of the following: OCR text extraction verification, digital signature and timestamp verification, and document hash verification.

3. The dual-qualified teacher recommendation system based on big data analysis according to claim 1, characterized in that, The mandatory items for teacher characteristic assessment include educational background and years of work experience. The educational background requires a master's degree or above and the graduating institution must be registered with the Ministry of Education. The years of work experience must be at least a minimum.

4. The dual-qualified teacher recommendation system based on big data analysis according to claim 3, characterized in that, In the teacher characteristic assessment, the bottleneck value is the minimum value of teaching experience, number of training projects and trainee feedback scores after normalization. Based on this, teachers are divided into four competency levels: Level 1, Level 2, Level 3 and Level 4.

5. The dual-qualified teacher recommendation system based on big data analysis according to claim 1, characterized in that, In the dual-key hard-door screening of matching and filtering, the professional relevance is calculated by using the TF-IDF model to calculate the cosine similarity between course requirements and teacher background, and the practical years need to be verified by checking the project contract to confirm that the cumulative years meet the standard.

6. The dual-qualified teacher recommendation system based on big data analysis according to claim 5, characterized in that, The Pareto skyline matching filter constructs a feature vector based on years of teaching experience, student feedback ratings, and the number of training programs, eliminating dominated, disadvantaged candidates.

7. A dual-qualified teacher recommendation system based on big data analysis according to claim 5, characterized in that, The matching and filtering process prioritizes comparing the average score of student feedback based on lexicographical order, followed by comparing the number of training programs and years of teaching experience.

8. The dual-qualified teacher recommendation system based on big data analysis according to claim 1, characterized in that, In constraint control, reliability, coverage, and identifiability are calculated for the four major modules of layer A, the sub-modules of layer B, and the sub-items of layer C, and a judgment matrix is ​​generated.

9. A dual-qualified teacher recommendation system based on big data analysis according to claim 8, characterized in that, In constraint control, the consistency of the matrix is ​​judged by the consistency index CI and the random consistency ratio CR. When CR < preset threshold, the normalized eigenvector is used as the weight.

10. A dual-qualified teacher recommendation system based on big data analysis according to claim 8, characterized in that, The weighted strategy of constraint control is only enabled when all decision matrices meet the consistency requirement; otherwise, it automatically reverts to the default strategy and outputs the decision results of the first three modules.

Images