Financial course multi-dimensional evaluation method and system based on RPA behavior analysis

Abstract

The invention discloses a financial course multi-dimensional evaluation method and system based on RPA behavior analysis, and relates to the technical field of education evaluation, and the method comprises the steps: collecting student learning behavior data and student attribute data from a plurality of key learning scenes through an RPA robot; the method comprises the steps of constructing a financial knowledge graph tree, constructing four first-level evaluation dimensions and a plurality of second-level indexes of the four first-level evaluation dimensions according to student learning behavior data, student attribute data and the financial knowledge graph tree, obtaining the student type of each student, determining a dynamic influence factor based on the student type and the student attribute data, constructing a three-dimensional influence matrix of each secondary index based on the dynamic influence factor, obtaining a dynamic weight of each secondary index, obtaining a comprehensive capability value of each primary evaluation dimension based on the dynamic weight of each secondary index, and constructing a financial course learning radar map of each student in the current collection period; the course recommendation list is generated according to the financial course learning radar map, so that the financial course evaluation result is more objective and accurate.

Description

Technical Field

[0001] The present invention relates to the field of educational evaluation technology, and in particular to a multi-dimensional evaluation method and system for financial courses based on RPA behavior analysis. Background Art

[0002] A Chinese patent with publication number CN116823028A discloses a teaching quality evaluation system and method, including an information collection module and a comprehensive evaluation module; the information collection module is communicatively connected to the comprehensive evaluation module; the information collection module is used to collect multidimensional evaluation parameters corresponding to each course taught by a target teacher; the multidimensional evaluation parameters include student evaluation parameters, teacher evaluation parameters, as well as student classroom participation indicators, student attendance indicators and student performance indicators; the comprehensive evaluation module is used to input the multidimensional evaluation parameters into a comprehensive evaluation model to obtain the teaching quality evaluation results corresponding to each course, and to integrate the teaching quality evaluation results corresponding to all courses taught by the target teacher to obtain the comprehensive teaching quality score result of the target teacher.

[0003] Chinese patent publication number CN116562703A discloses a multi-dimensional joint scoring course evaluation method with corrected matching, including the following steps: S1: students are divided into groups to complete team reports and conduct mutual evaluation based on the completion status; S2: the entropy weight-Topsis combined method is used to give the CI value of the inter-group evaluation, a mapping interval is established according to the score, and the CI value is projected onto the assignment space of [80,100]; S3: a mapping interval is also established for mutual evaluation and self-evaluation, and the score is projected onto the assignment space of [80,100]; S4: weighted assignment is performed on the practical cooperation part; S5: a scoring revision system is established, the course assessment part is assigned, the difficulty coefficient is verified and adjusted; S6: the results after the practice + assessment scoring are weighted to obtain the final evaluation of the student for the course, and the normality test of the final evaluation of the course is performed.

[0004] Traditional evaluation methods often rely on limited data sources, making it difficult to comprehensively collect students' learning behavior data and student attribute data in various key learning scenarios, resulting in a one-sided understanding of students' learning status. The evaluation dimensions are single, and it is impossible to comprehensively consider students' abilities from multiple dimensions such as knowledge understanding, practical operation skills, professional ethics, learning strategies and efficiency. The weight setting is usually static and fixed, and cannot be flexibly adjusted according to dynamic factors such as students' course progress, student type and industry goals. It is difficult to accurately reflect students' true ability levels at different stages and goals. At the same time, due to the lack of effective data mining and analysis methods, it is impossible to deeply analyze students' mastery of financial knowledge by constructing a knowledge graph tree. It is also difficult to obtain key indicators reflecting the learning process, such as learning strategies corresponding to operation trigger paths and overall cross-layer jump efficiency. As a result, it is impossible to provide students with scientific and personalized course recommendations, which is not conducive to meeting students' diverse learning needs and the precise training of financial professionals. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention aims to provide a multi-dimensional evaluation method for financial courses based on RPA behavior analysis, which includes the following steps:

[0006] like Figure 1 As shown in the figure, the multi-dimensional evaluation method of financial courses based on RPA behavior analysis includes the following steps:

[0007] Step s1: Use RPA robots to automatically collect student learning behavior data and student attribute data from multiple sources in several key learning scenarios (including financial practice software platforms, learning management systems, and collaborative learning platforms) and mark the collection period;

[0008] Step s2: Construct a financial knowledge graph tree. Based on students' learning behavior data and student attribute data, obtain students' knowledge mastery of each node in the financial knowledge graph tree, knowledge node coverage, learning strategies corresponding to operation trigger paths, overall cross-layer jump efficiency of operation trigger paths, process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient.

[0009] Step s3: Construct four first-level evaluation dimensions and several second-level indicators of the four first-level evaluation dimensions, screen several second-level indicators as evaluation indicators, obtain the student type of each student, and determine the dynamic impact factor based on the student type and student attribute data;

[0010] Step s4: Based on the dynamic impact factors, a three-dimensional impact matrix is ​​constructed for each secondary indicator, and the dynamic weight of each secondary indicator is obtained. Based on the dynamic weight of each secondary indicator, the comprehensive ability value of each primary evaluation dimension is obtained, and a financial course learning radar chart for each student in the current collection period is constructed;

[0011] Step s5: Generate a course recommendation list based on the finance course learning radar chart of each student in the current collection period.

[0012] Furthermore, student learning behavior data includes learning resource browsing records, financial operation process records, and collaborative learning process records, and student attribute data includes course progress and industry goals.

[0013] Furthermore, a financial knowledge graph tree is constructed, and the students' knowledge mastery of each node in the financial knowledge graph tree is obtained based on their learning behavior data. The process of knowledge node coverage includes:

[0014] Acquire multi-source financial knowledge data in advance, perform entity extraction and relationship extraction on the multi-source financial knowledge data, obtain multiple entities and the connection relationships between multiple entities, define attributes for each entity, preset a three-tier architecture, and classify multiple entities into a three-tier architecture. The three-tier architecture includes a knowledge point layer, a skill point layer, and a professional ability layer. Use multiple entities as nodes and the connection relationships between multiple entities as the connection relationships between nodes to construct a financial knowledge graph tree.

[0015] Based on learning resource browsing records, financial operation process records, and the financial knowledge graph tree, obtain the student's stay time at each node in the knowledge point layer, the practical task error rate associated with each node, and the operation trigger path;

[0016] A quantification rule table is preset, which includes the knowledge mastery quantification degree corresponding to different residence times and practical task error rates. The knowledge mastery degree of each node in the knowledge point layer is obtained based on the residence time of the node, the practical task error rate associated with the node, and the quantification rule table;

[0017] Preset the knowledge mastery threshold, compare the knowledge mastery of each node in the knowledge point layer with the knowledge mastery threshold, mark the nodes whose knowledge mastery is greater than the knowledge mastery threshold as learned nodes, obtain the total number of nodes included in the current course progress, and obtain the knowledge node coverage based on the number of learned nodes and the total number of nodes.

[0018] Furthermore, the process of obtaining the learning strategy corresponding to the operation trigger path and the overall cross-layer jump efficiency of the operation trigger path includes:

[0019] Obtain the architecture, access time, browsing order, dwell time, and error rate of each node in the operation trigger path, and preset the judgment conditions corresponding to different learning strategies.

[0020] The browsing order, dwell time, and practical task error rate of each node included in the operation trigger path are compared with the judgment conditions corresponding to different learning strategies to obtain the learning strategy corresponding to the operation trigger path. At the same time, the overall cross-layer jump efficiency of the operation trigger path is obtained based on the architecture, browsing order, access time, dwell time, and practical task error rate of each node.

[0021] It should be further explained that, in a specific implementation, the process of obtaining the overall cross-layer jump efficiency of the operation trigger path includes:

[0022] Let n i is the i-th node, containing attributes (l i ,t i ,s i )(level, visit time, length of stay), D i,j For slave node n i To node n j Jump (must satisfy (j=i+1), that is, adjacent browsing node), Δl i,j is the jump level difference, Δl=l j -l i , jump upward Δl=+1, jump downward Δl=-1, jump on the same layer Δl=0, T i,j is the jump time interval, T i,j =t j -(t i +s i )(the interval between the end time of the previous node and the start time of the next node), E r is the error rate of the practical task (the value range is 0-1, 0 means no error, 1 means all errors);

[0023]

[0024] Among them, SJE i,j represents the single jump efficiency index between node i and node j, PCLE is the overall cross-layer jump efficiency of the operation trigger path, |Δl i,j | represents the "effective level span" of the jump between nodes i and j, μ is the minimum value, and the correction term To reflect the influence of the previous node’s stay time, S max is the reasonable maximum stay time of the node at this level. Staying too long may lead to efficiency degradation. m represents all cross-layer jump events in the operation triggering path. SJE k Represents the single jump efficiency index of the kth cross-layer jump event, and calculates the SJE of each jump k And sum it up, take the average and multiply it by the error rate correction factor (the higher the error rate, the lower the overall efficiency).

[0025] Furthermore, the process of obtaining process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient includes:

[0026] Preset a standard financial process, compare the financial operation process records with the standard financial process, obtain the number of compliant steps, the total number of standard process steps, the number of data errors, the total number of data entry times, the actual total time spent by students, the total time spent on the standard process, the number of redundant steps, and the total number of student operation steps; obtain the process compliance rate (CR), data accuracy rate (DA), process efficiency index (EI), and redundant operation rate (RR) based on the number of compliant steps, the total number of standard process steps, the number of data errors, the total number of data entry times, the actual total time spent by students, the total time spent on the standard process, the number of redundant steps, and the total number of student operation steps;

[0027] in,

[0028]

[0029]

[0030] EI>100% indicates higher efficiency (shorter time consumption), while EI<100% indicates lower efficiency;

[0031] During the practical application of financial software, RPA is used to accurately record students' voucher entry steps, report generation logic, and data verification processes. By comparing these with standard financial processes, it automatically identifies operational deviations (such as debit and credit imbalances and process redundancies), and quantitatively assesses the standardization and efficiency of practical skills, thus filling the gap in traditional practical evaluations that are "results-oriented and ignore process."

[0032] The coefficients of various indicators in the collaborative learning process records were extracted. These coefficients included the number of times abnormal financial data was discovered in case analysis and actual operations, the proportion of effective communication in group work, the contribution to team goals, the proportion of tasks completed on time, and the rigor with which tasks were approached. The sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient were obtained based on each type of indicator.

[0033] The calculation formulas for the sensitivity coefficient (SC), communication and collaboration coefficient (CC), and responsibility coefficient (RC) are as follows:

[0034]

[0035] Among them, N a The number of abnormal data that are actively marked, such as uneven credit and debit balances, missing voucher attachments, abnormal transaction amounts, etc.; N t is the total number of data checks; δ is the conversion coefficient, C eTo increase the number of effective communications, the content involves discussions on financial rules, resolution of data discrepancies, and optimization of task division. RPA captures and identifies keywords in the discussion area, such as "applicable standards" and "process optimization." t is the total number of communications, including all collaborative behaviors such as questions, replies, and file sharing; T i The quality of task completion for individuals (such as compliance rate of voucher entry, accuracy rate of report analysis, with a value of 0-1); i is the task weight (defined according to the complexity of the financial task, such as "consolidated report preparation" (V = 0.9), "data collection" (V = 0.5)); W c 、W g is the weight of communication and contribution; R t = Number of tasks submitted on time / total number of tasks; E r is the number of low-level errors that occurred during the task (such as digital entry errors and format errors, which are automatically detected by financial software); E t The total number of task operations (such as the number of voucher entry fields, the number of report formula setting items); t 、W p Weighted for on-time completion and rigor.

[0036] Furthermore, the four first-level evaluation dimensions include knowledge understanding, practical skills, professionalism, and learning strategies and efficiency.

[0037] Several secondary indicators in the knowledge understanding dimension include the knowledge node coverage rate of the course progress and the knowledge mastery of each node. Several secondary indicators in the practical operation skills dimension include process compliance rate, data accuracy rate, process efficiency index, and redundant operation rate. Several secondary indicators in the professional quality dimension include sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient. Several secondary indicators in the learning strategy and efficiency dimension include learning strategy and overall cross-level jump efficiency.

[0038] Furthermore, the process of screening several secondary indicators as evaluation indicators and obtaining the student type of each student includes:

[0039] Several secondary indicators of the knowledge understanding dimension and several secondary indicators of the learning strategy and efficiency dimension of each student are used as evaluation indicators, and the indicator weights of the evaluation indicators and the student type fuzzy rule base are preset. The student type fuzzy rule base includes the threshold ranges of the evaluation indicators corresponding to different student types. Based on the student type fuzzy rule base, the membership matrix of each student to different student types is obtained through fuzzy comprehensive evaluation, and the student type of each student is obtained according to the membership matrix and the indicator weights.

[0040] Furthermore, a three-dimensional influence matrix of each secondary indicator is constructed based on the dynamic impact factor, and the dynamic weight of each secondary indicator is obtained. Based on the dynamic weight of each secondary indicator, the comprehensive ability value of each primary evaluation dimension is obtained. The process of constructing a financial course learning radar chart for each student in the current collection cycle includes:

[0041] Mark students' course progress, student type, and industry goals as dynamic influencing factors, collect learning behavior data and student attribute data of several students within the historical collection cycle, perform correlation analysis on the learning behavior data and student attribute data of several students within the historical collection cycle, and construct a three-dimensional influence matrix for each secondary indicator based on the correlation analysis results. The three-dimensional influence matrix includes adjustment coefficients for the secondary indicators based on the transformation relationships of different dynamic influencing factors;

[0042] The process of performing correlation analysis on the historical learning behavior data and historical student attribute data of several students includes:

[0043] Randomly select a historical collection cycle from several students' historical collection cycles and mark it as a sample cycle; mark the course progress, student type and industry target in the sample cycle as sample course progress, sample student type and sample industry target; select a historical collection cycle from several students' historical collection cycles whose course progress, student type and industry target are not completely consistent with the sample course progress, sample student type and sample industry target; perform Pearson correlation coefficient analysis on the secondary indicators in the historical collection cycle and the secondary indicators in the sample cycle to obtain the Pearson correlation coefficient between the secondary indicators; quantify the Pearson correlation coefficient into an adjustment coefficient; and simultaneously obtain the dynamic impact factors of the historical collection cycle and the sample cycle; obtain a dynamic impact factor transformation relationship between the historical collection cycle and the sample cycle, wherein the dynamic impact factor transformation relationship includes the dynamic impact factor of the historical collection cycle and the dynamic impact factor of the sample cycle; and mark the adjustment coefficient as the adjustment coefficient of the dynamic impact factor transformation relationship for the secondary indicator;

[0044] Repeat the above process to obtain the adjustment coefficients of several different dynamic impact factor transformation relationships on the secondary indicators;

[0045] The calculation formula for quantifying the Pearson correlation coefficient into the adjustment coefficient is:

[0046] α=1+k·r;

[0047] Where α is the adjustment coefficient, r is the Pearson correlation coefficient, r∈[-1,+1], and k is the degree scaling factor, which is set according to the Pearson correlation coefficient: |r| ≥ 0.7: k = 0.5 (strong positive enhancement by 50%, strong negative reduction by 50%), 0.3 ≤ |r| < 0.7: k = 0.3 (medium adjustment by 30%), and |r| < 0.3: k = 0.1 (weak adjustment by 10%).

[0048] Obtain the dynamic influence factor of the student's previous collection cycle and the dynamic weight of each secondary indicator at the end timestamp of the previous collection cycle, construct a dynamic influence factor transformation relationship based on the dynamic influence factor of the student's current collection cycle and the dynamic influence factor of the previous collection cycle, obtain the adjustment coefficient of the dynamic influence factor corresponding to each secondary indicator based on the three-dimensional influence matrix and the dynamic influence factor transformation relationship, use the dynamic weight of each secondary indicator at the end timestamp of the previous collection cycle as the initial weight of each secondary indicator in the current collection cycle, and obtain the dynamic weight of each secondary indicator based on the initial weight of each secondary indicator and the adjustment coefficient of the dynamic influence factor corresponding to each secondary indicator;

[0049] For each secondary indicator i, its dynamic weight is the product of the initial weight and the adjustment coefficient of the dynamic impact factor:

[0050] W i ′=W i ×(β T ·sc T +β S ·fu S +β I ·cr I );

[0051] Among them, W i ′ represents the dynamic weight of the secondary index i, β T , β S , β I are the importance weights of course progress, student type, and industry goals (determined by the Delphi method, where β T =0.4,β S =0.3,β I =0.3), sc T 、fu S 、cr I are the adjustment coefficients of course progress, student type, and industry goals on indicator i;

[0052] Perform weighted averaging on the secondary indicators included in each primary evaluation dimension in the current collection period according to the dynamic weight of each secondary indicator to obtain the comprehensive capability value of each primary evaluation dimension;

[0053] The four first-level evaluation dimensions are used as the four axes of the radar chart and arranged clockwise. The comprehensive ability value of each first-level evaluation dimension is converted into a radius value under polar coordinates to construct a financial course learning radar chart for students in the current collection period.

[0054] Furthermore, the process of generating a course recommendation list based on the finance course learning radar chart of each student in the current collection period includes:

[0055] According to the student's course progress and industry goals, obtain a standard financial course learning radar chart corresponding to the course progress and industry goals, compare the student's financial course learning radar chart with the standard financial course learning radar chart, and obtain the radius value deviation of each axis (axis radius value deviation = axis radius value of the standard financial course learning radar chart - axis radius value of the same position in the financial course learning radar chart);

[0056] Obtain the course relevance coefficients of several finance courses for different first-level evaluation dimensions. The course relevance coefficient indicates the degree of pertinence of a particular course in improving the first-level evaluation dimension and is determined by the degree of match between the course syllabus and the dimension. Its quantification method is to calculate the ability-enhancing effect of the course on each first-level evaluation dimension using historical learning data. For example, after studying the course, the average absolute value of the radius value deviation of the first-level evaluation dimension is increased (the radius value of the axis of the standard finance course learning radar chart corresponding to the first-level evaluation dimension is smaller than the radius value of the axis of the finance course learning radar chart corresponding to the first-level evaluation dimension) or the average absolute value of the radius value deviation of the first-level evaluation dimension is reduced (the radius value of the axis of the standard finance course learning radar chart corresponding to the first-level evaluation dimension is larger than the radius value of the axis of the finance course learning radar chart corresponding to the first-level evaluation dimension); based on the course relevance coefficients of each finance course for different first-level evaluation dimensions and the radius value deviation of the first-level evaluation dimension corresponding to each axis, obtain the course recommendation coefficient of each finance course;

[0057] The calculation formula for obtaining the course recommendation coefficient of each financial course is:

[0058] Let the knowledge understanding dimension be A, the practical operation skill dimension be B, the professional quality dimension be C, and the learning strategy and efficiency dimension be D;

[0059] Score K =∑ v∈{A,B,C,D} (D v ×R v,K );

[0060] Among them, Score K represents the course recommendation coefficient of finance course k, D v Indicates the radius value deviation of the first-level evaluation dimension v, R v,K represents the course correlation coefficient between the first-level evaluation dimension v and the financial course K;

[0061] Arrange the course recommendation coefficients of various financial courses from high to low to generate a course recommendation list.

[0062] A multi-dimensional evaluation system for financial courses based on RPA behavioral analysis includes a cloud, wherein the cloud is connected to a data acquisition module, a data processing module, an evaluation indicator construction module, a learning visualization module, and an evaluation recommendation module;

[0063] The data collection module uses RPA robots to automatically collect student learning behavior data and student attribute data from multiple sources in several key learning scenarios and mark the collection cycle;

[0064] The data processing module is used to construct a financial knowledge graph tree. Based on students' learning behavior data and student attribute data, it obtains students' knowledge mastery of each node in the financial knowledge graph tree, knowledge node coverage, learning strategies corresponding to operation trigger paths, overall cross-layer jump efficiency of operation trigger paths, process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient.

[0065] The evaluation index construction module is used to construct four first-level evaluation dimensions and several second-level indicators of the four first-level evaluation dimensions, screen several second-level indicators as evaluation indicators, obtain the student type of each student, and determine the dynamic impact factor based on the student type and student attribute data.

[0066] The learning visualization module is used to construct a three-dimensional influence matrix for each secondary indicator based on the dynamic impact factor, obtain the dynamic weight of each secondary indicator, obtain the comprehensive ability value of each primary evaluation dimension based on the dynamic weight of each secondary indicator, and construct a financial course learning radar chart for each student in the current collection cycle;

[0067] The evaluation and recommendation module is used to generate a course recommendation list based on the financial course learning radar chart of each student's current collection period.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] 1. Starting from the four first-level evaluation dimensions of knowledge understanding, practical operation skills, professional quality, learning strategy and efficiency, it covers multiple second-level indicators such as knowledge node coverage, knowledge mastery, process compliance rate, data accuracy, etc., to conduct a comprehensive and systematic evaluation of students' financial course learning, avoiding the one-sidedness of single-dimensional evaluation.

[0070] 2. Use RPA robots to collect student learning behavior data and student attribute data, including learning resource browsing records, financial operation process records, collaborative learning process records, course progress, industry goals, and other information, which can more comprehensively reflect students' learning status and characteristics.

[0071] 3. By analyzing and processing the actual collected data, such as calculating indicators such as knowledge mastery and process compliance based on objective data such as learning resource browsing records and financial operation process records, the interference of human factors is reduced, making the evaluation results more objective and accurate.

[0072] 4. Preset quantitative rule tables, knowledge mastery thresholds, etc., to quantify students' learning situation, can more accurately measure students' performance in various aspects, and provide a scientific basis for evaluation and course recommendations.

[0073] 5. Taking students' course progress, student types and industry goals as dynamic influencing factors, and constructing a three-dimensional influence matrix to obtain the dynamic weights of each secondary indicator, the weights of evaluation indicators can be dynamically adjusted according to the different stages and characteristics of students, reflecting the actual situation of students more accurately.

[0074] 6. Collect data from several students within the historical collection period for correlation analysis, and construct a three-dimensional influence matrix based on the analysis results, so that the evaluation model can adapt to changes in the students' learning process, adjust the evaluation and course recommendations in a timely manner, and meet the learning needs of students at different stages.

[0075] 7. Through fuzzy comprehensive evaluation, the membership matrix of each student to different student types is obtained, and then the student type is determined. Personalized evaluation and course recommendations can be provided for different types of students to improve learning outcomes.

[0076] 8. Based on the radius deviation between the student's financial course learning radar chart and the standard radar chart, as well as the correlation coefficient between the financial course and the courses in different first-level evaluation dimensions, the course recommendation coefficient is calculated and a course recommendation list is generated to provide each student with personalized course recommendations that are in line with their own learning situation and development goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 This is a schematic diagram of the multi-dimensional evaluation method for financial courses based on RPA behavioral analysis in an embodiment of the present application.

[0078] Figure 2 This is a schematic diagram of the multi-dimensional evaluation system for financial courses based on RPA behavioral analysis in an embodiment of the present application.

[0079] Figure 3 This is a financial knowledge graph tree architecture diagram of an embodiment of the present application. DETAILED DESCRIPTION

[0080] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0081] like Figure 1 As shown in the figure, the multi-dimensional evaluation method of financial courses based on RPA behavior analysis includes the following steps:

[0082] Step s1: Use RPA robots to automatically collect student learning behavior data and student attribute data from multiple sources in several key learning scenarios (including financial practice software platforms, learning management systems, and collaborative learning platforms) and mark the collection period;

[0083] Step s2: Construct a financial knowledge graph tree. Based on students' learning behavior data and student attribute data, obtain students' knowledge mastery of each node in the financial knowledge graph tree, knowledge node coverage, learning strategies corresponding to operation trigger paths, overall cross-layer jump efficiency of operation trigger paths, process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient.

[0084] Step s3: Construct four first-level evaluation dimensions and several second-level indicators of the four first-level evaluation dimensions, screen several second-level indicators as evaluation indicators, obtain the student type of each student, and determine the dynamic impact factor based on the student type and student attribute data;

[0085] Step s4: Based on the dynamic impact factors, a three-dimensional impact matrix is ​​constructed for each secondary indicator, and the dynamic weight of each secondary indicator is obtained. Based on the dynamic weight of each secondary indicator, the comprehensive ability value of each primary evaluation dimension is obtained, and a financial course learning radar chart for each student in the current collection period is constructed;

[0086] Step s5: Generate a course recommendation list based on the finance course learning radar chart of each student in the current collection period.

[0087] It should be further explained that, during the specific implementation process, student learning behavior data includes learning resource browsing records, financial operation process records, and collaborative learning process records, and student attribute data includes course progress and industry goals.

[0088] Using RPA robots, we automatically and non-invasively collect student learning behavior data from multiple key learning scenarios. In the learning management system, we capture data such as student login time, browsing time for learning resources (such as course videos and documents), and the order and frequency of access to each knowledge point. In the financial practice software platform, we record students' financial business operation processes in detail, including the operation time, data input content, operation error types, and correction records for each step, such as voucher filling, account book registration, and report preparation. In the collaborative learning platform, we collect data such as the division of labor among group members, the number of student speeches in the discussion forum, the content of speeches, and interactive communication records.

[0089] It should be further explained that, in the specific implementation process, the financial knowledge graph tree is constructed, and the students' knowledge mastery of each node in the financial knowledge graph tree is obtained based on their learning behavior data. The process of knowledge node coverage includes:

[0090] Acquire multi-source financial knowledge data in advance, perform entity extraction and relationship extraction on the multi-source financial knowledge data, obtain several entities and the connection relationships between several entities, and define the attributes of each entity. Preset a three-tier architecture and classify several entities into a three-tier architecture. The three-tier architecture includes a knowledge point layer: including 200+ atomic knowledge points such as "financial accounting concepts", "standard application", and "report preparation logic", a skill point layer: corresponding to 50+ practical skills such as "invoice review process" and "trial balance verification", and a professional ability layer: corresponding to core abilities such as "financial accounting ability" and "financial analysis ability". Take several entities as nodes and the connection relationships between several entities as the connection relationships between nodes to build a financial knowledge graph tree. Figure 3 This is a financial knowledge graph tree architecture diagram of an embodiment of the present application;

[0091] Based on learning resource browsing records, financial operation process records, and the financial knowledge graph tree, obtain the student's stay time at each node in the knowledge point layer, the practical task error rate associated with each node, and the operation trigger path;

[0092] A quantification rule table is preset, which includes the knowledge mastery quantification degree corresponding to different residence times and practical task error rates. The knowledge mastery degree of each node in the knowledge point layer is obtained based on the residence time of the node, the practical task error rate associated with the node, and the quantification rule table;

[0093] Preset a knowledge mastery threshold, compare the knowledge mastery of each node in the knowledge point layer with the knowledge mastery threshold, mark the nodes whose knowledge mastery is greater than the knowledge mastery threshold as learned nodes, obtain the total number of nodes included in the current course progress, and obtain the knowledge node coverage rate based on the number of learned nodes and the total number of nodes. Knowledge node coverage rate = number of learned nodes / total number of nodes.

[0094] It should be further explained that, during the specific implementation process, multi-source financial knowledge data covers the six core courses of "Basic Accounting", "Intermediate Financial Accounting", "Cost Accounting", "Financial Management", "Tax Law", and "Auditing", and extends to the emerging field of intelligent finance (such as RPA financial applications and business-finance integration systems). The multi-source financial knowledge data types include textbooks and outlines, accounting standards, practical operation processes, and professional competency standards.

[0095] Entities include: Knowledge point entities: "Accrual Basis of Accounting", "Inventory Impairment Provision", "Cost-Volume-Profit Analysis Model", Skill entities: "VAT Invoice Certification", "Financial Statement Cross-Reference Verification", "Budget Management System Operation", and Ability entities: "Financial Data Decision-Making Ability", "Tax Risk Identification Ability", and "Business and Financial Process Optimization Ability".

[0096] The connections between entities include: hierarchical relationships: "belongs to" (e.g., "Revenue Recognition" belongs to "Financial Accounting Capabilities"), supporting relationships: "supports" (e.g., "Debit-Entry Bookkeeping Knowledge Point" supports "Voucher Entry Skills"), association relationships: "association" (e.g., "Fixed Asset Depreciation Method" is associated with "Corporate Income Tax Adjustment"), and operational relationships: "operation steps" (e.g., "Prepare Balance Sheet" Step 1 → Step 2 → Step 3).

[0097] The attributes of the entity include: knowledge point attributes: difficulty level (L1-L5), update time (date of revision of the latest standards); skill attributes: practical time standard (such as the standard time of "VAT declaration" is 45 minutes), compliance checkpoints (such as "invoice verification requires checking the password area and amount").

[0098] It should be further explained that, in the specific implementation process, the process of obtaining the learning strategy corresponding to the operation trigger path and the overall cross-layer jump efficiency of the operation trigger path includes:

[0099] Obtain the architecture, access time, browsing order, dwell time, and practical task error rate of each node included in the operation trigger path, and preset judgment conditions corresponding to different learning strategies. The learning strategies include theory-first, trial-and-error, and mixed-cycle, for example:

[0100] Theory-first (T-type strategy): Judgment conditions: When first encountering a new module, the theoretical knowledge browsing time accounts for ≥ 60%, and practical tasks are not triggered before the basic theoretical learning is completed;

[0101] Example: When learning "Consolidated Financial Statements," first read through three theoretical videos (90 minutes total) before starting the simulation.

[0102] Practical trial and error (P-type strategy):

[0103] Judgment criteria: The error rate for the first practical operation is ≥ 40%, and the student immediately returns to the theoretical module to review the corresponding knowledge points after making an error (e.g., opening the "Lending and Borrowing Rules" explanation video within 10 minutes after a voucher entry error);

[0104] Example: When preparing a cash flow statement, try manually allocating items. After failing three times, review the "Indirect Method Preparation Steps" animation.

[0105] Mixed Cycle Type (TP Strategy):

[0106] Judgment criteria: The frequency of alternating visits between theory and practice is ≥ 3 rounds / class period (e.g., "10 minutes of theory → 5 minutes of practice → 5 minutes of reviewing key points → practice again");

[0107] The browsing order, dwell time, and practical task error rate of each node included in the operation trigger path are compared with the judgment conditions corresponding to different learning strategies to obtain the learning strategy corresponding to the operation trigger path. At the same time, the overall cross-layer jump efficiency of the operation trigger path is obtained based on the architecture, browsing order, access time, dwell time, and practical task error rate of each node.

[0108] It should be further explained that, in a specific implementation, the process of obtaining the overall cross-layer jump efficiency of the operation trigger path includes:

[0109] Let n i is the i-th node, containing attributes (l i ,t i ,s i )(level, visit time, length of stay), D i,j For slave node n i To node n j Jump (must satisfy (j=i+1), that is, adjacent browsing node), Δl i,j is the jump level difference, Δl=l j -l i , jump upward Δl=+1, jump downward Δl=-1, jump on the same layer Δl=0, T i,j is the jump time interval, T i,j =t j -(t i +s i )(the interval between the end time of the previous node and the start time of the next node), E r is the error rate of the practical task (the value range is 0-1, 0 means no error, 1 means all errors);

[0110]

[0111] Among them, SJE i,jrepresents the single jump efficiency index between node i and node j, PCLE is the overall cross-layer jump efficiency of the operation trigger path, |Δl i,j | represents the "effective level span" of the jump between nodes i and j, μ is the minimum value, and the correction term To reflect the influence of the previous node’s stay time, S max is the reasonable maximum stay time of the node at this level. Staying too long may lead to efficiency degradation. m represents all cross-layer jump events in the operation triggering path. SJE k Represents the single jump efficiency index of the kth cross-layer jump event, and calculates the SJE of each jump k And sum it up, take the average and multiply it by the error rate correction factor (the higher the error rate, the lower the overall efficiency).

[0112] It should be further explained that during the specific implementation process, the process of obtaining process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient includes:

[0113] Preset a standard financial process, compare the financial operation process records with the standard financial process, obtain the number of compliant steps, the total number of standard process steps, the number of data errors, the total number of data entry times, the actual total time spent by students, the total time spent on the standard process, the number of redundant steps, and the total number of student operation steps; obtain the process compliance rate (CR), data accuracy rate (DA), process efficiency index (EI), and redundant operation rate (RR) based on the number of compliant steps, the total number of standard process steps, the number of data errors, the total number of data entry times, the actual total time spent by students, the total time spent on the standard process, the number of redundant steps, and the total number of student operation steps;

[0114] in,

[0115]

[0116] EI>100% indicates higher efficiency (shorter time consumption), while EI<100% indicates lower efficiency;

[0117] During the practical application of financial software, RPA is used to accurately record students' voucher entry steps, report generation logic, and data verification processes. By comparing these with standard financial processes, it automatically identifies operational deviations (such as debit and credit imbalances and process redundancies), and quantitatively assesses the standardization and efficiency of practical skills, thus filling the gap in traditional practical evaluations that are "results-oriented and ignore process."

[0118] The coefficients of various indicators in the collaborative learning process records were extracted. These coefficients included the number of times abnormal financial data was discovered in case analysis and actual operations, the proportion of effective communication in group work, the contribution to team goals, the proportion of tasks completed on time, and the rigor with which tasks were approached. The sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient were obtained based on each type of indicator.

[0119] The calculation formulas for the sensitivity coefficient (SC), communication and collaboration coefficient (CC), and responsibility coefficient (RC) are as follows:

[0120]

[0121] Among them, N a The number of abnormal data that are actively marked, such as uneven credit and debit balances, missing voucher attachments, abnormal transaction amounts, etc.; N t is the total number of data checks; δ is the conversion coefficient, C e To increase the number of effective communications, the content involves discussions on financial rules, resolution of data discrepancies, and optimization of task division. RPA captures and identifies keywords in the discussion area, such as "applicable standards" and "process optimization." t is the total number of communications, including all collaborative behaviors such as questions, replies, and file sharing; T i The quality of task completion for individuals (such as compliance rate of voucher entry, accuracy rate of report analysis, with a value of 0-1); i is the task weight (defined according to the complexity of the financial task, such as "consolidated report preparation" (V = 0.9), "data collection" (V = 0.5)); W c 、W g is the weight of communication and contribution; R t = Number of tasks submitted on time / total number of tasks; E r is the number of low-level errors that occurred during the task (such as digital entry errors and format errors, which are automatically detected by financial software); E t The total number of task operations (such as the number of voucher entry fields, the number of report formula setting items); t 、W p Weighted for on-time completion and rigor.

[0122] It should be further explained that, in the specific implementation process, the four first-level evaluation dimensions include knowledge understanding dimension, practical operation skill dimension, professional quality dimension and learning strategy and efficiency dimension;

[0123] Several secondary indicators in the knowledge understanding dimension include the knowledge node coverage rate of the course progress and the knowledge mastery of each node. Several secondary indicators in the practical operation skills dimension include process compliance rate, data accuracy rate, process efficiency index, and redundant operation rate. Several secondary indicators in the professional quality dimension include sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient. Several secondary indicators in the learning strategy and efficiency dimension include learning strategy and overall cross-level jump efficiency.

[0124] Traditional financial course evaluation focuses on "knowledge mastery" (such as test scores), while this indicator system constructs a "four-dimensional ability matrix" for the first time, for example:

[0125] Deepen vertical development: Set financial-specific indicators such as "process compliance rate" and "process efficiency index" in the "practical operation skills dimension" (such as the proportion of voucher entry steps that comply with accounting standards, and the average time to identify and correct credit and debit imbalances), and convert standards such as the "Enterprise Accounting Standards" and "Accounting Basic Work Standards" into quantifiable process indicators to solve the problem of traditional evaluation that "emphasizes results but ignores process."

[0126] Horizontal expansion: New indicators such as "data sensitivity" and "communication and collaboration ability coefficient" have been added to the "professional quality dimension" (such as the number of times abnormal transactions are actively marked in case analysis and the proportion of effective financial process optimization suggestions in group work). These indicators are directly connected to the core capabilities of financial positions (such as risk identification in auditing and cross-departmental collaboration in business and finance integration), filling the gaps in the traditional evaluation of "financial professional awareness".

[0127] It should be further explained that, in the specific implementation process, the process of screening several secondary indicators as evaluation indicators and obtaining the student type of each student includes:

[0128] Several secondary indicators of the knowledge understanding dimension and several secondary indicators of the learning strategy and efficiency dimension of each student are used as evaluation indicators, and the indicator weights of the evaluation indicators and the student type fuzzy rule base are preset. The student type fuzzy rule base includes the threshold ranges of the evaluation indicators corresponding to different student types. Based on the student type fuzzy rule base, the membership matrix of each student to different student types is obtained through fuzzy comprehensive evaluation, and the student type of each student is obtained according to the membership matrix and the indicator weights.

[0129] It should be further explained that, in the specific implementation process, the process of obtaining the student type of each student based on the membership matrix and the indicator weight matrix includes:

[0130] The indicator weight matrix and the membership matrix of the evaluation indicators are integrated through a formula to obtain a fuzzy comprehensive evaluation matrix of the evaluation indicators. The membership of each student to different student types is obtained according to the fuzzy comprehensive evaluation matrix. The student type with the highest membership corresponding to each student is screened out, and the student type with the highest membership corresponding to each student is used as the student type of each student.

[0131] Wherein, the formula is:

[0132] M = αM1 × βM2;

[0133] Among them, M is the fuzzy comprehensive evaluation matrix of the evaluation index, M1 is the indicator weight matrix of the evaluation index, M2 is the membership matrix, "×" represents the multiplication of the elements at corresponding positions of the weight matrix of the evaluation index and the membership matrix, and α and β are weighting parameters used to control the balance between the weight matrix and the membership matrix in the fuzzy comprehensive evaluation matrix of the evaluation index.

[0134] The student types include weak foundation type (S1):

[0135] Weight tilt: learning strategy and efficiency dimension (25% to 30%);

[0136] Data basis: clustering of students whose knowledge node coverage rate in the learning platform is less than 30% and whose knowledge mastery of each node is less than 30%;

[0137] Practice-oriented (S2):

[0138] Weighting tilt: Professional quality dimension (35%-40%), such as "data sensitivity" and "collaborative contribution";

[0139] Data basis: Clustering of students whose "process compliance rate > 90%" but "number of effective group communication times < 50% of the average" in the practical system;

[0140] Balanced development type (S3):

[0141] Weight balance: Dynamic floating of 20% to 30% in four dimensions;

[0142] Data basis: Student groups that do not belong to S1 and S2;

[0143] It should be further explained that, in the specific implementation process, the three-dimensional influence matrix of each secondary indicator is constructed based on the dynamic impact factor, the dynamic weight of each secondary indicator is obtained, and the comprehensive ability value of each primary evaluation dimension is obtained based on the dynamic weight of each secondary indicator. The process of constructing the financial course learning radar chart for each student in the current collection cycle includes:

[0144] Mark students' course progress, student type, and industry goals as dynamic influencing factors, collect learning behavior data and student attribute data of several students within the historical collection cycle, perform correlation analysis on the learning behavior data and student attribute data of several students within the historical collection cycle, and construct a three-dimensional influence matrix for each secondary indicator based on the correlation analysis results. The three-dimensional influence matrix includes adjustment coefficients for the secondary indicators based on the transformation relationships of different dynamic influencing factors;

[0145] The process of performing correlation analysis on the historical learning behavior data and historical student attribute data of several students includes:

[0146] Randomly select a historical collection cycle from several students' historical collection cycles and mark it as a sample cycle; mark the course progress, student type and industry target in the sample cycle as sample course progress, sample student type and sample industry target; select a historical collection cycle from several students' historical collection cycles whose course progress, student type and industry target are not completely consistent with the sample course progress, sample student type and sample industry target; perform Pearson correlation coefficient analysis on the secondary indicators in the historical collection cycle and the secondary indicators in the sample cycle to obtain the Pearson correlation coefficient between the secondary indicators; quantify the Pearson correlation coefficient into an adjustment coefficient; and simultaneously obtain the dynamic impact factors of the historical collection cycle and the sample cycle; obtain a dynamic impact factor transformation relationship between the historical collection cycle and the sample cycle, wherein the dynamic impact factor transformation relationship includes the dynamic impact factor of the historical collection cycle and the dynamic impact factor of the sample cycle; and mark the adjustment coefficient as the adjustment coefficient of the dynamic impact factor transformation relationship for the secondary indicator;

[0147] Repeat the above process to obtain the adjustment coefficients of several different dynamic impact factor transformation relationships on the secondary indicators;

[0148] The calculation formula for quantifying the Pearson correlation coefficient into the adjustment coefficient is:

[0149] α=1+k·r;

[0150] Where α is the adjustment coefficient, r is the Pearson correlation coefficient, r∈[-1,+1], and k is the degree scaling factor, which is set according to the Pearson correlation coefficient: |r| ≥ 0.7: k = 0.5 (strong positive enhancement by 50%, strong negative reduction by 50%), 0.3 ≤ |r| < 0.7: k = 0.3 (medium adjustment by 30%), and |r| < 0.3: k = 0.1 (weak adjustment by 10%).

[0151] Obtain the dynamic influence factor of the student's previous collection cycle and the dynamic weight of each secondary indicator at the end timestamp of the previous collection cycle, construct a dynamic influence factor transformation relationship based on the dynamic influence factor of the student's current collection cycle and the dynamic influence factor of the previous collection cycle, obtain the adjustment coefficient of the dynamic influence factor corresponding to each secondary indicator based on the three-dimensional influence matrix and the dynamic influence factor transformation relationship, use the dynamic weight of each secondary indicator at the end timestamp of the previous collection cycle as the initial weight of each secondary indicator in the current collection cycle, and obtain the dynamic weight of each secondary indicator based on the initial weight of each secondary indicator and the adjustment coefficient of the dynamic influence factor corresponding to each secondary indicator;

[0152] For each secondary indicator i, its dynamic weight is the product of the initial weight and the adjustment coefficient of the dynamic impact factor:

[0153] W i ′=W i ×(β T ·sc T +β S ·fu S +β I ·cr I );

[0154] Among them, W i ′ represents the dynamic weight of the secondary index i, β T , β S , β I are the importance weights of course progress, student type, and industry goals (determined by the Delphi method, where β T =0.4,β S =0.3,β I =0.3,),sc T 、fu S 、cr I are the adjustment coefficients of course progress, student type, and industry goals on indicator i;

[0155] Perform weighted averaging on the secondary indicators included in each primary evaluation dimension in the current collection period according to the dynamic weight of each secondary indicator to obtain the comprehensive capability value of each primary evaluation dimension;

[0156] The four first-level evaluation dimensions are used as the four axes of the radar chart and arranged clockwise. The comprehensive ability value of each first-level evaluation dimension is converted into a radius value under polar coordinates to construct a financial course learning radar chart for students in the current collection period.

[0157] It should be further explained that, in the specific implementation process, the process of generating a course recommendation list based on the financial course learning radar chart of each student's current collection period includes:

[0158] According to the student's course progress and industry goals, obtain a standard financial course learning radar chart corresponding to the course progress and industry goals, compare the student's financial course learning radar chart with the standard financial course learning radar chart, and obtain the radius value deviation of each axis (axis radius value deviation = axis radius value of the standard financial course learning radar chart - axis radius value of the same position in the financial course learning radar chart);

[0159] Obtain the course relevance coefficients of several finance courses for different first-level evaluation dimensions. The course relevance coefficient indicates the degree of pertinence of a particular course in improving the first-level evaluation dimension and is determined by the degree of match between the course syllabus and the dimension. Its quantification method is to calculate the ability-enhancing effect of the course on each first-level evaluation dimension using historical learning data. For example, after studying the course, the average absolute value of the radius value deviation of the first-level evaluation dimension is increased (the radius value of the axis of the standard finance course learning radar chart corresponding to the first-level evaluation dimension is smaller than the radius value of the axis of the finance course learning radar chart corresponding to the first-level evaluation dimension) or the average absolute value of the radius value deviation of the first-level evaluation dimension is reduced (the radius value of the axis of the standard finance course learning radar chart corresponding to the first-level evaluation dimension is larger than the radius value of the axis of the finance course learning radar chart corresponding to the first-level evaluation dimension); based on the course relevance coefficients of each finance course for different first-level evaluation dimensions and the radius value deviation of the first-level evaluation dimension corresponding to each axis, obtain the course recommendation coefficient of each finance course;

[0160] The calculation formula for obtaining the course recommendation coefficient of each financial course is:

[0161] Let the knowledge understanding dimension be A, the practical operation skill dimension be B, the professional quality dimension be C, and the learning strategy and efficiency dimension be D;

[0162] Score K =Σ v∈{A,B,C,D} (D v ×R v,K );

[0163] Among them, Score K represents the course recommendation coefficient of finance course k, D v Indicates the radius value deviation of the first-level evaluation dimension v, R v,K represents the course correlation coefficient between the first-level evaluation dimension v and the financial course K;

[0164] Arrange the course recommendation coefficients of various financial courses from high to low to generate a course recommendation list.

[0165] like Figure 2 As shown, a multi-dimensional evaluation system for financial courses based on RPA behavior analysis includes a cloud, wherein the cloud is connected to a data acquisition module, a data processing module, an evaluation index construction module, a learning visualization module, and an evaluation recommendation module;

[0166] The data collection module uses RPA robots to automatically collect student learning behavior data and student attribute data from multiple sources in several key learning scenarios and mark the collection cycle;

[0167] The data processing module is used to construct a financial knowledge graph tree. Based on students' learning behavior data and student attribute data, it obtains students' knowledge mastery of each node in the financial knowledge graph tree, knowledge node coverage, learning strategies corresponding to operation trigger paths, overall cross-layer jump efficiency of operation trigger paths, process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient.

[0168] The evaluation index construction module is used to construct four first-level evaluation dimensions and several second-level indicators of the four first-level evaluation dimensions, screen several second-level indicators as evaluation indicators, obtain the student type of each student, and determine the dynamic impact factor based on the student type and student attribute data.

[0169] The learning visualization module is used to construct a three-dimensional influence matrix for each secondary indicator based on the dynamic impact factor, obtain the dynamic weight of each secondary indicator, obtain the comprehensive ability value of each primary evaluation dimension based on the dynamic weight of each secondary indicator, and construct a financial course learning radar chart for each student in the current collection cycle;

[0170] The evaluation and recommendation module is used to generate a course recommendation list based on the financial course learning radar chart of each student's current collection period.

[0171] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A multi-dimensional evaluation method for financial courses based on RPA behavioral analysis, characterized by: The following steps are involved: Step s1: Use RPA robots to collect student learning behavior data and student attribute data from several key learning scenarios and mark the collection period; Step s2: Construct a financial knowledge graph tree. Based on the student learning behavior data, student attribute data, and the financial knowledge graph tree, obtain the student's knowledge mastery of each node in the financial knowledge graph tree, knowledge node coverage, learning strategy corresponding to the operation trigger path, overall cross-layer jump efficiency of the operation trigger path, process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient; Step s3: Construct four first-level evaluation dimensions and several second-level indicators of the four first-level evaluation dimensions, screen several second-level indicators as evaluation indicators, obtain the student type of each student, and determine the dynamic impact factor based on the student type and student attribute data; Step s4: Based on the dynamic impact factors, a three-dimensional impact matrix is ​​constructed for each secondary indicator, and the dynamic weight of each secondary indicator is obtained. Based on the dynamic weight of each secondary indicator, the comprehensive ability value of each primary evaluation dimension is obtained, and a financial course learning radar chart for each student in the current collection period is constructed; Step s5: Generate a course recommendation list based on the finance course learning radar chart.

2. The multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to claim 1 is characterized in that: Student learning behavior data includes learning resource browsing records, financial operation process records, and collaborative learning process records; student attribute data includes course progress and industry goals.

3. The multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to claim 2 is characterized in that: Construct a financial knowledge graph tree and obtain students' knowledge mastery of each node in the financial knowledge graph tree based on students' learning behavior data. The process of knowledge node coverage includes: Acquire multi-source financial knowledge data in advance, perform entity extraction and relationship extraction on the multi-source financial knowledge data, obtain multiple entities and the connection relationships between multiple entities, preset a three-tier architecture, classify multiple entities into a three-tier architecture, and use the three-tier architecture to classify multiple entities. The three-tier architecture includes a knowledge point layer, a skill point layer, and a professional ability layer. Use multiple entities as nodes and the connection relationships between multiple entities as the connection relationships between nodes to build a financial knowledge graph tree. Based on learning resource browsing records, financial operation process records, and the financial knowledge graph tree, obtain the student's stay time at each node in the knowledge point layer, the practical task error rate associated with each node, and the operation trigger path; A preset quantitative rule table is used to obtain the knowledge mastery of each node in the knowledge point layer based on the node's residence time, the error rate of the practical task associated with the node, and the quantitative rule table; A knowledge mastery threshold is preset, and nodes whose knowledge mastery is greater than the knowledge mastery threshold are marked as learned nodes. The total number of nodes included in the current course progress is obtained, and the knowledge node coverage is obtained based on the number of learned nodes and the total number of nodes.

4. The multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to claim 3 is characterized in that: The process of obtaining the learning strategy corresponding to the operation trigger path and the overall cross-layer jump efficiency of the operation trigger path includes: Obtain the architecture, access time, browsing order, dwell time, and operational task error rate of each node in the operation trigger path; Preset judgment conditions corresponding to different learning strategies, compare the browsing order, dwell time, and practical task error rate of each node included in the operation trigger path with the judgment conditions corresponding to different learning strategies, and obtain the learning strategy corresponding to the operation trigger path; Based on the architecture, browsing order, access time, dwell time, and actual task error rate of each node, the overall cross-layer jump efficiency of the operation trigger path is obtained.

5. The multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to claim 4 is characterized in that: The process of obtaining process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient includes: Preset standard financial processes, compare financial operation process records with standard financial processes, and obtain process compliance rate, data accuracy rate, process efficiency index and redundant operation rate; Extract the coefficients of various types of indicators in the collaborative learning process records, and obtain the sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient based on each type of indicator.

6. The multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to claim 5 is characterized in that: The four first-level evaluation dimensions include knowledge understanding, practical skills, professionalism, and learning strategies and efficiency. Several secondary indicators of the knowledge understanding dimension include the knowledge node coverage rate of the course progress and the knowledge mastery of each node; several secondary indicators of the practical operation skills dimension include process compliance rate, data accuracy rate, process efficiency index, and redundant operation rate; several secondary indicators of the professional quality dimension include sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient; several secondary indicators of the learning strategy and efficiency dimension include learning strategy and overall cross-level jump efficiency.

7. The multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to claim 6 is characterized in that: The process of selecting several secondary indicators as evaluation indicators and obtaining the student type of each student includes: Several secondary indicators of the knowledge understanding dimension and several secondary indicators of the learning strategy and efficiency dimension of each student are used as evaluation indicators, and the indicator weights of the evaluation indicators and the student type fuzzy rule base are preset. The student type fuzzy rule base includes the threshold ranges of the evaluation indicators corresponding to different student types. Based on the student type fuzzy rule base, the membership matrix of each student to different student types is obtained through fuzzy comprehensive evaluation, and the student type of each student is obtained according to the membership matrix and the indicator weights.

8. The multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to claim 7 is characterized in that: The process of constructing a three-dimensional influence matrix for each secondary indicator based on the dynamic impact factor, obtaining the dynamic weight of each secondary indicator, and obtaining the comprehensive ability value of each primary evaluation dimension based on the dynamic weight of each secondary indicator, and constructing a financial course learning radar chart for each student in the current collection cycle includes: Mark students' course progress, student type, and industry goals as dynamic influencing factors, collect learning behavior data and student attribute data of several students within the historical collection cycle, perform correlation analysis on the learning behavior data and student attribute data of several students within the historical collection cycle, and construct a three-dimensional influence matrix for each secondary indicator based on the correlation analysis results. The three-dimensional influence matrix includes adjustment coefficients for the secondary indicators based on the transformation relationships of different dynamic influencing factors; Obtain the dynamic influence factor of the student's previous collection cycle and the dynamic weight of each secondary indicator at the end timestamp of the previous collection cycle, construct a dynamic influence factor transformation relationship based on the dynamic influence factor of the student's current collection cycle and the dynamic influence factor of the previous collection cycle, obtain the adjustment coefficient of the dynamic influence factor corresponding to each secondary indicator based on the three-dimensional influence matrix and the dynamic influence factor transformation relationship, use the dynamic weight of each secondary indicator at the end timestamp of the previous collection cycle as the initial weight of each secondary indicator in the current collection cycle, and obtain the dynamic weight of each secondary indicator based on the initial weight of each secondary indicator and the adjustment coefficient of the dynamic influence factor corresponding to each secondary indicator; Perform weighted averaging on the secondary indicators included in each primary evaluation dimension according to the dynamic weight of each secondary indicator to obtain the comprehensive capability value of each primary evaluation dimension; The four first-level evaluation dimensions are used as the four axes of the radar chart and arranged clockwise. The comprehensive ability value of each first-level evaluation dimension is converted into a radius value under polar coordinates to construct a financial course learning radar chart.

9. The multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to claim 8 is characterized in that: The process of generating a course recommendation list based on the finance course learning radar chart of each student's current collection period includes: According to the student's course progress and industry goals, obtain a standard financial course learning radar chart corresponding to the course progress and industry goals, compare the student's financial course learning radar chart with the standard financial course learning radar chart, and obtain the radius value deviation of each axis; Obtain the course relevance coefficients of several financial courses for different first-level evaluation dimensions, and obtain the course recommendation coefficient of each financial course based on the course relevance coefficients of each financial course for different first-level evaluation dimensions and the radius value deviation of the first-level evaluation dimension corresponding to each axis; Arrange the course recommendation coefficients of various financial courses from high to low to generate a course recommendation list.

10. A multi-dimensional evaluation system for financial courses based on RPA behavior analysis, specifically applied to the multi-dimensional evaluation method for financial courses based on RPA behavior analysis according to any one of claims 1 to 9, characterized in that: The cloud comprises a data acquisition module, a data processing module, an evaluation index construction module, a learning visualization module and an evaluation recommendation module. The data collection module uses RPA robots to automatically collect student learning behavior data and student attribute data from multiple sources in several key learning scenarios and mark the collection cycle; The data processing module is used to construct a financial knowledge graph tree. Based on students' learning behavior data and student attribute data, it obtains students' knowledge mastery of each node in the financial knowledge graph tree, knowledge node coverage, learning strategies corresponding to operation trigger paths, overall cross-layer jump efficiency of operation trigger paths, process compliance rate, data accuracy rate, process efficiency index, redundant operation rate, sensitivity coefficient, communication and collaboration ability coefficient, and responsibility coefficient. The evaluation index construction module is used to construct four primary evaluation dimensions for each student and several secondary indicators of the four primary evaluation dimensions, screen several secondary indicators as evaluation indicators, obtain the student type of each student, and determine the dynamic impact factor based on the student type and student attribute data. The learning visualization module is used to construct a three-dimensional influence matrix for each secondary indicator based on the dynamic impact factor, obtain the dynamic weight of each secondary indicator, obtain the comprehensive ability value of each primary evaluation dimension based on the dynamic weight of each secondary indicator, and construct a financial course learning radar chart for each student in the current collection cycle; The evaluation and recommendation module is used to generate a course recommendation list based on the financial course learning radar chart of each student's current collection period.

Images