Customer demand intelligent matching management method and system based on big data

Abstract

The invention discloses a customer demand intelligent matching management method and system based on big data, and belongs to the technical field of online education. The system comprises a data acquisition module, a demand analysis module, an intelligent matching module and a display management module. The data acquisition module acquires behavior data when a client object browses a teaching video, and analyzes the contained learning object and browsing record; the demand analysis module analyzes the browsing record, establishes a cognitive library, analyzes knowledge points contained in the teaching video, constructs a relation tree diagram, analyzes behavior data, maps the knowledge points in the relation tree diagram, and calculates a mastering index; the intelligent matching module is used for establishing a recommendation library and positioning in the relation tree diagram; calculating a weight coefficient of each knowledge point and a matching index of each teaching video in the recommendation library according to the positioning combination mastering index; and the display management module sets a dynamic recommendation list according to the change of the matching index, and when the client object exits browsing of the teaching video, freeze-frame the recommendation list and display and push the recommendation list.

Description

Technical Field

[0001] This invention relates to the field of online education technology, specifically to a method and system for intelligent matching and management of customer needs based on big data. Background Technology

[0002] With the rapid development of internet technology and mobile communications, online education has become an important supplement and transformative force in education supply. Platform-based, fragmented learning and the concept of lifelong learning are driving a shift in educational services from a "course-centered" to a "learner-centered" approach. At the same time, users exhibit significant differences in learning objectives, knowledge base, and learning habits. How to efficiently match massive educational resources with diverse user needs has become a core issue for improving teaching effectiveness, user retention, and commercialization.

[0003] Currently, online education platforms face a structural contradiction between the large-scale expansion of teaching resources and the deepening of users' personalized needs. Especially in video learning scenarios, existing recommendation architectures have significant limitations: First, the connection between viewing behavior features and the inherent semantic units of the video is disconnected; relying solely on discrete behavioral tags cannot establish a spatiotemporal cognitive continuity map, causing subsequent learning analysis to deviate from the true cognitive state. Second, dynamic modeling capabilities are severely lacking; static processing modes are slow to respond to the heterogeneous evolutionary characteristics of time-series sequences, making it difficult to capture the fractal growth patterns of individual learning paths. Finally, the push mechanism exhibits a mechanical, haphazard approach, and universal strategies cannot adapt to the phase transition needs of individual cognitive states. This rigid technological paradigm causes recommendation services to remain at the level of coarse resource screening for a long time, making it difficult to achieve truly intelligent learning path navigation. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for intelligent matching and management of customer needs based on big data, so as to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides a customer demand intelligent matching and management method based on big data, comprising:

[0006] S100: Collect behavioral data of customer users browsing instructional videos, as well as the attribute parameters of the instructional videos. Parse the learning objects contained in the attribute parameters, and the browsing history of each learning object.

[0007] The target audience refers to users who are watching the instructional videos. Instructional videos are digital educational resources that use video as a medium to present course content.

[0008] Behavioral data refers to the feedback actions that users take at different times while watching instructional videos. Feedback actions include submitting questions and answers, adjusting speed, and jumping to different progress.

[0009] The attribute parameters include all learners who have watched the corresponding instructional video. The browsing history includes all instructional videos viewed by the learner.

[0010] Build a structured data pool of user behavior and resource attributes to provide underlying support for subsequent analysis.

[0011] S200: Analyze browsing history to filter teaching videos and establish a cognitive library and an extended library. Analyze the knowledge points contained in each teaching video and construct a relationship tree diagram. Analyze behavioral data to map knowledge points in the relationship tree diagram and calculate the mastery index. Specifically, this includes:

[0012] S201. Establish a cognitive database by extracting all instructional videos from the customer's browsing history and adding them to the cognitive database. Count the number of instructional videos in the cognitive database. And the number of all learners included in each instructional video. .

[0013] S202. Establish an extended library, analyze the teaching videos in the browsing history of each learner, and extract these... All instructional videos for each learner have been added to the extended library. .

[0014] Due to computational constraints, the number of instructional videos in the extended library needs to be controlled. This can be achieved by filtering instructional videos from browsing history and by filtering the learning targets of each instructional video in the cognitive library, thus limiting the number of instructional videos in the extended library.

[0015] S203. Analyze the knowledge points and time periods contained in each teaching video in the cognitive and extended libraries, analyze the chronological order of different knowledge points, establish the progressive relationship between knowledge points, and thus construct a relationship tree diagram. Specifically, this includes:

[0016] S2031. Analyze all teaching videos in the cognitive library and extended library, use a speech recognition engine to convert the speech in the teaching videos into text, and use OCR technology to recognize images in the teaching videos and extract text, and merge the text from the same time period.

[0017] It adopts dual-modal fusion technology of speech recognition + OCR to break through the limitations of a single information source.

[0018] S2032. Perform semantic segmentation on the fused text, use natural language understanding algorithms to identify semantic conceptual entities, establish logical relationship structure between conceptual entities in combination with contextual relationships, and generate a set of knowledge points.

[0019] S2033. Synchronously parse the timestamp records of the video stream, take the start and end time nodes of each knowledge point in the video as the time period of its implementation, and output the final time sequence mapping each knowledge point to its corresponding time period.

[0020] By dividing the boundaries of knowledge points through time sequence, precise slicing of knowledge points at the granular level can be achieved.

[0021] S2034. Arrange all knowledge points under each teaching video according to the time sequence of each segment, and construct knowledge chains for each. Analyze the relationships between knowledge points based on all knowledge chains and establish progressive relationships, then construct a relationship tree diagram based on these progressive relationships. Specifically, this includes:

[0022] S2034-1 Obtaining Instructional Videos For all knowledge points in the corresponding knowledge chain, arrange each pair of adjacent knowledge points in chronological order. and Divided into combinations, knowledge points within each combination and They serve as the parent node and child node, respectively.

[0023] S2034-2. Divide the knowledge chain corresponding to each teaching video into groups. Divide all groups with the same two knowledge points into the same category. Calculate the proportion of the two types of groups in the same category where the parent node and child node have the same knowledge points.

[0024] S2034-3. Mark classes where the proportion of any combination exceeds a preset threshold, and analyze the knowledge points corresponding to the parent and child nodes of the high-proportion combinations in the marked classes. and Setting up knowledge points arrive The progressive relationship.

[0025] S2034-4, for unlabeled classes, analyze two knowledge points in the combination and establish a peer-level relationship between them. Obtain all knowledge points in all knowledge chains and construct a relationship tree diagram based on the progressive or peer-level relationships between the knowledge points.

[0026] S204. Extract behavioral data from customers browsing each instructional video in the knowledge base, analyze the time period of each feedback action corresponding to the knowledge point, and calculate the mastery index of each knowledge point, mapping them in a relational tree diagram. Specifically, this includes:

[0027] S2041, Extracting Customer Objects and Browsing Tutorial Videos behavioral data Analyze teaching videos Includes time periods for each knowledge point, as well as behavioral data. The time period corresponding to each feedback action.

[0028] S2042. Calculate the accuracy rate of the questions and answers for each knowledge point. Percentage of unplayed frames Percentage of high-speed playback time Percentage of frames played per session Question and answer error rate Percentage of frames already played Percentage of low-speed playback time and the percentage of frames played multiple times Specifically, this includes:

[0029] S2042-1. Analyze the knowledge points in the time period corresponding to the submission of the question and answer instruction, and take the percentage of correct answers and the percentage of incorrect answers as the correct answer rate and the incorrect answer rate of the corresponding knowledge points, respectively.

[0030] S2042-2, Analyze the progress position and adjustment value corresponding to the speed adjustment command, as shown in the teaching video. The timeline is divided into intervals with different playback speeds, and the video playback speed remains constant within the same interval.

[0031] S2042-3, Preset Normal Playback Speed Playback speed greater than As a high-speed range, the playback speed is less than... As a low-speed range.

[0032] S2042-4. Divide the total duration of the high-speed interval and the total duration of the low-speed interval in each time period by the duration of the time period to obtain the percentage of high-speed playback time and the percentage of low-speed playback time for the corresponding knowledge points.

[0033] S2042-5, Analysis of Teaching Videos Includes image frames; analyzes jump progress instructions in the teaching video. The jump-out and jump-in points correspond to the image frames, as well as whether each image frame is played and the number of times it is played.

[0034] S2042-6. Count the total number of image frames for each knowledge point within the time period. Divide the number of unplayed image frames and the number of played image frames within the time period by the total number of image frames to obtain the percentage of unplayed frames and the percentage of played frames for the corresponding knowledge point.

[0035] S2042-7. Divide the number of single-play image frames and the number of multiple-play image frames within the time period by the total number of image frames to obtain the percentage of single-play frames and the percentage of multiple-play frames for the corresponding knowledge points.

[0036] Using as the input variable, negative feedback is passed through the exponential term. Achieve nonlinear amplification.

[0037] The sigmoid function is used to simulate the saturation characteristics of the learning curve during the teaching process.

[0038] S2043, Preset curve steepness factor Positive feedback gain coefficient Negative feedback suppression coefficient and the sharpness of perception of difficulties Tutorial video on substituting formulas for calculation Mastery coefficient of each knowledge point :

[0039] ;

[0040] Positive feedback gain coefficient Automatically adjusts based on course type. Negative feedback inhibition coefficient. Used in the dynamic response learning phase.

[0041] The parameters are automatically adjusted according to the teaching scenario to avoid mechanical deviations from general formulas.

[0042] Sharpness of difficulty perception The power amplification effect used for negative feedback, through The parameters achieve the "learning bottleneck amplification effect," which aligns with the difficulty sensitivity characteristics in the teaching process.

[0043] Curve steepness factor Used to control the sensitivity of the evaluation.

[0044] in, and These are the positive feedback factor and the negative feedback factor, respectively, and their calculation formulas are as follows:

[0045] ;

[0046] ;

[0047] In the formula, .

[0048] Introducing an asymmetric feedback mechanism, positive feedback Including metrics such as question-and-answer accuracy, negative feedback pass >1 Implement "error penalty amplification".

[0049] The mastery coefficient formula achieves a non-linear quantitative representation of the degree of knowledge mastery through the sigmoid function transformation mechanism. Its core input consists of the difference between the positive feedback factor and the negative feedback factor.

[0050] Among them, the positive feedback factor integrates positive behavioral indicators such as the correct answer rate and the proportion of unplayed frames, while the negative feedback factor integrates negative behavioral characteristics such as the error rate of question and answer and the frequency of repeated playback frames.

[0051] The adaptive gain adjustment factor embedded in the formula dynamically adjusts the reinforcement of positive feedback based on the course attributes, while the inhibition factor controls the weakening of negative feedback according to the characteristics of the learning stage.

[0052] By introducing a difficulty perception sharpness parameter, a nonlinear amplification effect is applied to negative feedback, effectively simulating the sensitive response characteristics to learning bottlenecks in the teaching and cognitive process.

[0053] The curve steepness control parameter is used to modulate the discriminative sensitivity of the evaluation results, ensuring sufficient gradient change resolution within the critical capability threshold range.

[0054] S2044. Following this pattern, calculate the mastery coefficient for each knowledge point under each instructional video in the cognitive base. Calculate the average mastery coefficient for the same knowledge point across all instructional videos in the cognitive base as the mastery index. .

[0055] S2045. Map all the mastery indices of all knowledge points to a relational tree diagram.

[0056] Establish topological relationships between knowledge points and quantify users' mastery levels to achieve a digital mapping of cognitive states.

[0057] S300. Build a recommendation library based on the instructional videos currently being viewed by the target audience and locate them in a relationship tree diagram. Calculate the weight coefficient of each knowledge point and the matching index of each instructional video in the recommendation library based on the location and the mastery index. Specifically, this includes:

[0058] S301. Obtain the tutorial videos currently being viewed in the customer's browsing history. The videos are then labeled with all the knowledge points they contain. A recommendation library is created, and all instructional videos from the extended library that contain at least one labeled knowledge point are added to the recommendation library.

[0059] S302, Analyzing Teaching Videos The current playback progress is in a specific time period. The corresponding knowledge point in this time period is marked as a location point. The location point is mapped in the relationship tree diagram, and the area where the location point and all its subordinate knowledge points are located is divided as a reference area.

[0060] The reference area is formed by radiating outwards from the knowledge point corresponding to the current playback progress.

[0061] Sub-level knowledge points include all knowledge points below the positioning point, and each knowledge point can form a knowledge point propagation chain with the positioning point.

[0062] S303. Within the reference area, starting from the positioning point and ending at knowledge points without subordinates, different relationship chains are planned. All knowledge points in each relationship chain are interconnected according to hierarchical relationships, and the knowledge points in different relationship chains are not completely identical.

[0063] All knowledge points in a relationship chain are within the reference area, containing at least one location point and one knowledge point without subordinates, and each relationship chain contains at least two knowledge points.

[0064] S304. Take all knowledge points within the reference area, excluding the positioning points, as reference points, and count the number of knowledge points containing reference points. Number of all relationship chains Obtain reference points Mastery Index Calculate the reference point Weighting coefficients :

[0065] ;

[0066] In the formula, To include reference points The The number of all reference points in the relationship chain. For the first Reference point in the direction from the start to the end of the relationship chain The ordinal number. Calculate the weight coefficients of each other reference point within the reference area.

[0067] The item assigns greater weight to the front end of the chain, ensuring that weak points in the foundation are strengthened first. This allows knowledge points with low mastery to receive higher recommendation priority.

[0068] The weighting coefficient formula constructs a dynamic weighting mechanism based on a knowledge topology network, determining the recommendation priority of specific knowledge points through multi-path aggregation analysis. During the calculation process, all relational chain paths containing the target knowledge point are enumerated, and calculations are performed on each path.

[0069] The application of path weight allocation factors enhances the decision-making influence of core links. This factor is jointly determined by the total number of path nodes and the target node order parameter, thereby achieving a weight allocation that tilts towards basic nodes in the composite knowledge chain.

[0070] A reverse mapping mechanism for knowledge point mastery is introduced, using the complement of the mastery index as an adjustment coefficient to ensure that knowledge points with low mastery receive higher recommendation weights.

[0071] Multi-path normalization is performed by taking the inverse of the number of paths to eliminate weight bias caused by differences in path density.

[0072] The weighting coefficients integrate the structural characteristics of the knowledge system with the differences in learners' cognitive states to form a topology-aware priority determination algorithm.

[0073] S305. Analyze the knowledge points and corresponding time periods contained in each teaching video in the recommendation library, and calculate the matching index for each teaching video by substituting them into the formula. :

[0074] ;

[0075] In the formula, The number of reference points contained in the instructional video. The first video in the teaching series The duration of each reference point corresponds to a specific time period. The first video in the teaching series The weighting coefficients of each reference point This refers to the duration of the instructional video.

[0076] Matching Index The weighted knowledge points correspond to the time period percentage in the calculation ( This avoids briefly mentioned knowledge points from having an excessive impact on the recommendation results.

[0077] A matching evaluation model for teaching resources and knowledge needs is constructed, with its basic calculation paradigm being a time-weighted integration of each knowledge dimension. All reference knowledge points covered by the teaching videos being evaluated are selected.

[0078] For a single knowledge point, the length of its presentation time in the video is coupled with the corresponding weight value to generate a local matching degree, and global normalization is performed by taking the reciprocal of the total time.

[0079] The local matching degrees of all reference knowledge points are linearly superimposed to form the final matching index. By introducing a time-weighted mechanism, the depth of knowledge point teaching is accurately represented, while the dynamic injection of weight coefficients achieves adaptive matching with the learner's cognitive state.

[0080] Generate a personalized recommendation strategy based on prioritizing cognitive weaknesses.

[0081] S400: Analyze the changes in the matching index of each instructional video and set a dynamic recommendation list. When a user exits the browsing of instructional videos, freeze the recommendation list and push it to the user. Update behavioral data and attribute parameters based on browsing behavior. Specifically, this includes:

[0082] S401. Calculate the matching index of each teaching video in the recommendation library in real time, and select the top videos in descending order of matching index. Each instructional video is set up with a recommendation list, which is updated in real time based on changes in the matching index.

[0083] S402. When a client exits the tutorial video. When browsing, stop updating and freeze all teaching videos in the recommended list, then push them to the client's browsing interface for display.

[0084] Freeze the list when exiting the browser to solve the flickering interference problem caused by traditional real-time recommendations and preserve the integrity of the user's intent.

[0085] S403, Update customer browsing history and browsing tutorial videos. Real-time behavioral data is used to update instructional videos after using customer subjects as learning targets. The attribute parameters.

[0086] Achieve a closed-loop dynamic recommendation system without interference and enable the algorithm to evolve autonomously.

[0087] The present invention also provides a customer demand intelligent matching management system based on big data, including a data acquisition module, a demand analysis module, an intelligent matching module, and a display management module.

[0088] The data acquisition module collects behavioral data of customers browsing instructional videos and analyzes the learning objects and browsing history contained in the instructional videos.

[0089] Collect behavioral data generated by customers while browsing instructional videos, including specific timestamp feedback actions such as submitting questions and answers, adjusting speed, and jumping to the next step.

[0090] Simultaneously, the attribute parameters of the teaching videos are analyzed, including the video's metadata items and corresponding browsing history. This data acquisition relies on direct parsing of the video stream and timestamps, ensuring the real-time nature and completeness of the data collection.

[0091] It provides accurate basic data support, helps the system build a personalized initial data pool by capturing users' real-time interactive behavior, thereby improving the accuracy of subsequent analysis and recommendation modules, avoiding deviations caused by missing data, and ultimately achieving accurate construction of user profiles.

[0092] The requirements analysis module analyzes browsing history and builds a knowledge base, parses the knowledge points contained in teaching videos and constructs a relationship tree diagram, analyzes behavioral data to map knowledge points in the relationship tree diagram and calculates the mastery index.

[0093] Based on the collected browsing history, instructional videos are selected to establish a cognitive library and an extended library.

[0094] The video content is then analyzed, and the text is extracted and semantics are integrated using a speech recognition engine and OCR technology. Natural language processing algorithms are then used to segment semantic segments, identify conceptual entities, and construct progressive relationships between knowledge points to form a relational tree diagram.

[0095] At the same time, a knowledge point mastery index is calculated, and variables are used to quantify the degree of mastery.

[0096] It achieves deep modeling of user needs, accurately maps users' cognitive weaknesses through structured knowledge point dependencies and adaptive parameters, improves the system's ability to identify learning progress and difficulties, provides quantifiable evidence for personalized recommendations, and significantly enhances users' learning efficiency.

[0097] The intelligent matching module is used to build a recommendation library and locate content within a relational tree diagram. Based on this location and the mastery index, it calculates the weight coefficient of each knowledge point and the matching index of each teaching video in the recommendation library.

[0098] A recommendation library is built based on the videos currently viewed by the user, knowledge point areas are located in the relationship tree diagram, and relationship chains from the starting point to the end point are planned.

[0099] The weight of each knowledge point is calculated using a weighting coefficient formula, and then substituted into the matching index formula to evaluate the matching degree of videos in the recommendation library, dynamically adjusting the recommendation priority.

[0100] By optimizing the recommendation logic using algorithms, taking into account the coherence of the learning path and the weight allocation of the mastery index, personalized video suggestions are dynamically generated, effectively improving the real-time performance and relevance of recommendations, reducing user search costs, and optimizing the learning experience.

[0101] The display management module sets a dynamic recommendation list based on changes in the matching index. When a customer exits the browsing of the teaching videos, the recommendation list is frozen and displayed.

[0102] The display management module calculates the matching index of videos in the recommendation library in real time and dynamically generates a recommendation list. Updates stop when the user exits the browser to freeze the list, which is then pushed to the user interface. Simultaneously, the user's browsing history and video attribute parameters are updated.

[0103] Ensure smooth delivery of instructional videos and responsive user interface, and improve user experience by pausing the recommendation list at key moments to avoid distractions.

[0104] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0105] Behavioral semantic deep mapping: This solution achieves spatiotemporal alignment and analysis of user learning behaviors (such as progress jumps and playback speed adjustments) with video knowledge points, accurately capturing the cognitive value of micro-interactions. This avoids the state modeling bias caused by shallow feature extraction in traditional technologies, thereby improving the accuracy of user demand recognition.

[0106] Dynamic knowledge topology construction: The solution designs a dynamic modeling mechanism based on logical progression, which integrates discrete knowledge points into a coherent tree-like relationship chain, overcoming the cognitive discontinuity problem that is easily caused by the existing static labeling system, and ensuring the continuity of the learning path and the effectiveness of knowledge progression.

[0107] Real-time scenario response capability: The solution introduces an adaptive adjustment mechanism to optimize the recommendation priority enhancement strategy in real time for different course characteristics (such as theoretical and practical), which solves the lag and homogeneity of traditional systems in dynamic adjustment and significantly enhances the scenario adaptability of recommended content.

[0108] Data closed-loop optimization mechanism: By constructing a continuously iterative feedback loop, the system can instantly integrate new behavioral data to feed back into the knowledge graph update, eliminating the risk of recommendation performance decay caused by traditional unidirectional data flow, and achieving long-term stable evolution of the recommendation system and maintenance of user stickiness. Attached Figure Description

[0109] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0110] Figure 1 This is a flowchart illustrating the intelligent matching and management method for customer needs based on big data, as described in this invention.

[0111] Figure 2 This is a schematic diagram of the structure of the intelligent matching management system for customer needs based on big data, which is a present invention. Detailed Implementation

[0112] 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.

[0113] Example 1: Please refer to Figure 1 This invention provides a customer demand intelligent matching management method based on big data, including:

[0114] S100: Collect behavioral data of customer users browsing instructional videos, as well as the attribute parameters of the instructional videos. Parse the learning objects contained in the attribute parameters, and the browsing history of each learning object.

[0115] The target audience refers to users who are watching the instructional videos. Instructional videos are digital educational resources that use video as a medium to present course content.

[0116] Behavioral data refers to the feedback actions that users take at different times while watching instructional videos. Feedback actions include submitting questions and answers, adjusting speed, and jumping to different progress.

[0117] The attribute parameters include all learners who have watched the corresponding instructional video. The browsing history includes all instructional videos viewed by the learner.

[0118] In the specific implementation process, a structured data pool of user behavior and resource attributes is built to provide underlying support for subsequent analysis.

[0119] S200: Analyze browsing history to filter teaching videos and establish a cognitive library and an extended library. Analyze the knowledge points contained in each teaching video and construct a relationship tree diagram. Analyze behavioral data to map knowledge points in the relationship tree diagram and calculate the mastery index. Specifically, this includes:

[0120] S201. Establish a cognitive database by extracting all instructional videos from the customer's browsing history and adding them to the cognitive database. Count the number of instructional videos in the cognitive database. And the number of all learners included in each instructional video. .

[0121] S202. Establish an extended library, analyze the teaching videos in the browsing history of each learner, and extract these... All instructional videos for each learner have been added to the extended library. .

[0122] In the specific implementation process, due to computing power considerations, the number of teaching videos in the extended library needs to be controlled. This is achieved by filtering teaching videos in the browsing history to limit the number (e.g., filtering a certain number of teaching videos according to the viewing time sequence) and filtering the learning objects of each teaching video in the cognitive library to limit the number (e.g., filtering a certain number of learning objects according to the viewing time sequence), thus achieving strict control over the number of teaching videos in the extended library.

[0123] S203. Analyze the knowledge points and time periods contained in each teaching video in the cognitive and extended libraries, analyze the chronological order of different knowledge points, establish the progressive relationship between knowledge points, and thus construct a relationship tree diagram. Specifically, this includes:

[0124] S2031. Analyze all teaching videos in the cognitive library and extended library, use a speech recognition engine to convert the speech in the teaching videos into text, and use OCR technology to recognize images in the teaching videos and extract text, and merge the text from the same time period.

[0125] In the specific implementation process, the dual-modal fusion technology of speech recognition + OCR is adopted to overcome the limitations of a single information source (such as relying on speech to restore content when video subtitles are missing).

[0126] S2032. Perform semantic segmentation on the fused text, use natural language understanding algorithms to identify semantic conceptual entities, establish logical relationship structure between conceptual entities in combination with contextual relationships, and generate a set of knowledge points.

[0127] S2033. Synchronously parse the timestamp records of the video stream, take the start and end time nodes of each knowledge point in the video as the time period of its implementation, and output the final time sequence mapping each knowledge point to its corresponding time period.

[0128] In the specific implementation process, the boundaries of knowledge points are divided by time sequence to achieve precise slicing of knowledge point granularity (e.g., the time period for "definition of derivative" is 00:05-00:12).

[0129] S2034. Arrange all knowledge points under each teaching video according to the time sequence of each segment, and construct knowledge chains for each. Analyze the relationships between knowledge points based on all knowledge chains and establish progressive relationships, then construct a relationship tree diagram based on these progressive relationships. Specifically, this includes:

[0130] S2034-1 Obtaining Instructional Videos For all knowledge points in the corresponding knowledge chain, arrange each pair of adjacent knowledge points in chronological order. and Divided into combinations, knowledge points within each combination and They serve as the parent node and child node, respectively.

[0131] S2034-2. Divide the knowledge chain corresponding to each teaching video into groups. Divide all groups with the same two knowledge points into the same category. Calculate the proportion of the two types of groups in the same category where the parent node and child node have the same knowledge points.

[0132] S2034-3. Mark classes where the proportion of any combination exceeds a preset threshold, and analyze the knowledge points corresponding to the parent and child nodes of the high-proportion combinations in the marked classes. and Setting up knowledge points arrive The progressive relationship.

[0133] S2034-4, for unlabeled classes, analyze two knowledge points in the combination and establish a peer-level relationship between them. Obtain all knowledge points in all knowledge chains and construct a relationship tree diagram based on the progressive or peer-level relationships between the knowledge points.

[0134] S204. Extract behavioral data from customers browsing each instructional video in the knowledge base, analyze the time period of each feedback action corresponding to the knowledge point, and calculate the mastery index of each knowledge point, mapping them in a relational tree diagram. Specifically, this includes:

[0135] S2041, Extracting Customer Objects and Browsing Tutorial Videos behavioral data Analyze teaching videos Includes time periods for each knowledge point, as well as behavioral data. The time period corresponding to each feedback action.

[0136] S2042. Calculate the accuracy rate of the questions and answers for each knowledge point. Percentage of unplayed frames Percentage of high-speed playback time Percentage of frames played per session Question and answer error rate Percentage of frames already played Percentage of low-speed playback time and the percentage of frames played multiple times Specifically, this includes:

[0137] S2042-1. Analyze the knowledge points in the time period corresponding to the submission of the question and answer instruction, and take the percentage of correct answers and the percentage of incorrect answers as the correct answer rate and the incorrect answer rate of the corresponding knowledge points, respectively.

[0138] S2042-2, Analyze the progress position and adjustment value corresponding to the speed adjustment command, as shown in the teaching video. The timeline is divided into intervals with different playback speeds, and the video playback speed remains constant within the same interval.

[0139] S2042-3, Preset Normal Playback Speed Playback speed greater than As a high-speed range, the playback speed is less than... As a low-speed range.

[0140] S2042-4. Divide the total duration of the high-speed interval and the total duration of the low-speed interval in each time period by the duration of the time period to obtain the percentage of high-speed playback time and the percentage of low-speed playback time for the corresponding knowledge points.

[0141] S2042-5, Analysis of Teaching Videos Includes image frames; analyzes jump progress instructions in the teaching video. The jump-out and jump-in points correspond to the image frames, as well as whether each image frame is played and the number of times it is played.

[0142] S2042-6. Count the total number of image frames for each knowledge point within the time period. Divide the number of unplayed image frames and the number of played image frames within the time period by the total number of image frames to obtain the percentage of unplayed frames and the percentage of played frames for the corresponding knowledge point.

[0143] S2042-7. Divide the number of single-play image frames and the number of multiple-play image frames within the time period by the total number of image frames to obtain the percentage of single-play frames and the percentage of multiple-play frames for the corresponding knowledge points.

[0144] In the specific implementation process, the (exponential decay effect of positive feedback and negative feedback) is used as the input variable, and the negative feedback is transmitted through the exponential term. Achieve nonlinear amplification (rapidly lower the mastery coefficient when there are severe learning disabilities).

[0145] The sigmoid function is used to simulate the saturation characteristics of the learning curve during the teaching process (the 0.5 dividing point is set to quantify the degree of knowledge mastery).

[0146] S2043, Preset curve steepness factor Positive feedback gain coefficient Negative feedback suppression coefficient and the sharpness of perception of difficulties Tutorial video on substituting formulas for calculation Mastery coefficient of each knowledge point :

[0147] ;

[0148] Positive feedback gain coefficient Automatically adjusts based on course type (e.g., theory course) =1.1, Practical Course =0.9). Negative feedback suppression coefficient Used in dynamic response learning phases (such as new lessons) =1.8, review lesson =1.2).

[0149] The parameters are automatically adjusted according to the teaching scenario to avoid mechanical deviations from general formulas.

[0150] Sharpness of difficulty perception The power amplification effect used for negative feedback (e.g.) =2 indicates a secondary penalty), through The parameters achieve the "learning bottleneck amplification effect," which aligns with the difficulty sensitivity characteristics in the teaching process.

[0151] Curve steepness factor Used to control evaluation sensitivity ( The larger the value, the higher the discrimination of the 0.3-0.7 range.

[0152] in, and These are the positive feedback factor and the negative feedback factor, respectively, and their calculation formulas are as follows:

[0153] ;

[0154] ;

[0155] In the formula, .

[0156] Introducing an asymmetric feedback mechanism, positive feedback Including metrics such as question-and-answer accuracy, negative feedback pass >1 Implement "error penalty amplification" (e.g.) When the value is 2, the error impact increases quadratically.

[0157] In practice, the mastery coefficient formula achieves a non-linear quantitative representation of the degree of knowledge mastery through the sigmoid function transformation mechanism. Its core input consists of the difference between the positive feedback factor and the negative feedback factor.

[0158] Among them, the positive feedback factor integrates positive behavioral indicators such as the correct answer rate and the proportion of unplayed frames, while the negative feedback factor integrates negative behavioral characteristics such as the error rate of question and answer and the frequency of repeated playback frames.

[0159] The adaptive gain adjustment factor embedded in the formula dynamically adjusts the reinforcement magnitude of positive feedback based on the course attributes (such as implementing gain enhancement for theoretical courses), while the inhibition factor regulates the weakening intensity of negative feedback according to the characteristics of the learning stage (such as enhancing the inhibition effect during the introduction of new knowledge points).

[0160] By introducing a difficulty perception sharpness parameter, a nonlinear amplification effect is applied to negative feedback, effectively simulating the sensitive response characteristics to learning bottlenecks in the teaching and cognitive process.

[0161] The curve steepness control parameter is used to modulate the discriminative sensitivity of the evaluation results, ensuring sufficient gradient change resolution within the critical capability threshold range.

[0162] S2044. Following this pattern, calculate the mastery coefficient for each knowledge point under each instructional video in the cognitive base. Calculate the average mastery coefficient for the same knowledge point across all instructional videos in the cognitive base as the mastery index. .

[0163] S2045. Map all the mastery indices of all knowledge points to a relational tree diagram.

[0164] In the specific implementation process, the topological relationship of knowledge points is established and the user's mastery level is quantified to achieve a digital mapping of cognitive state.

[0165] S300. Build a recommendation library based on the instructional videos currently being viewed by the target audience and locate them in a relationship tree diagram. Calculate the weight coefficient of each knowledge point and the matching index of each instructional video in the recommendation library based on the location and the mastery index. Specifically, this includes:

[0166] S301. Obtain the tutorial videos currently being viewed in the customer's browsing history. The videos are then labeled with all the knowledge points they contain. A recommendation library is created, and all instructional videos from the extended library that contain at least one labeled knowledge point are added to the recommendation library.

[0167] S302, Analyzing Teaching Videos The current playback progress is in a specific time period. The corresponding knowledge point in this time period is marked as a location point. The location point is mapped in the relationship tree diagram, and the area where the location point and all its subordinate knowledge points are located is divided as a reference area.

[0168] In the specific implementation process, the reference area is formed by taking the knowledge point corresponding to the current playback progress as the center and radiating its subordinate knowledge points (for example, when explaining "basics of integrals", the reference area covers subsequent knowledge points such as "application of definite integrals").

[0169] Sub-level knowledge points include all knowledge points below the positioning point, and each knowledge point can form a knowledge point propagation chain with the positioning point.

[0170] S303. Within the reference area, starting from the positioning point and ending at knowledge points without subordinates, different relationship chains are planned. All knowledge points in each relationship chain are interconnected according to hierarchical relationships, and the knowledge points in different relationship chains are not completely identical.

[0171] All knowledge points in a relationship chain are within the reference area, containing at least one location point and one knowledge point without subordinates, and each relationship chain contains at least two knowledge points.

[0172] S304. Take all knowledge points within the reference area, excluding the positioning points, as reference points, and count the number of knowledge points containing reference points. Number of all relationship chains Obtain reference points Mastery Index Calculate the reference point Weighting coefficients :

[0173] ;

[0174] In the formula, To include reference points The The number of all reference points in the relationship chain. For the first Reference point in the direction from the start to the end of the relationship chain The ordinal number. Calculate the weight coefficients of each other reference point within the reference area.

[0175] The item gives greater weight to the front end of the chain ( (The smaller the size, the closer it is to the starting point), ensuring that weak points in the foundation are strengthened first. This allows knowledge points with low mastery to receive higher recommendation priority.

[0176] The weighting coefficient formula constructs a dynamic weighting mechanism based on a knowledge topology network, determining the recommendation priority of specific knowledge points through multi-path aggregation analysis. During the calculation process, all relational chain paths containing the target knowledge point are enumerated, and calculations are performed on each path.

[0177] The application of path weight allocation factors enhances the decision-making influence of core links. This factor is jointly determined by the total number of path nodes and the target node order parameter, thereby achieving a weight allocation that tilts towards basic nodes in the composite knowledge chain.

[0178] A reverse mapping mechanism for knowledge point mastery is introduced, using the complement of the mastery index as an adjustment coefficient to ensure that knowledge points with low mastery receive higher recommendation weights.

[0179] Multi-path normalization is performed by taking the inverse of the number of paths to eliminate weight bias caused by differences in path density.

[0180] The weighting coefficients integrate the structural characteristics of the knowledge system with the differences in learners' cognitive states to form a topology-aware priority determination algorithm.

[0181] S305. Analyze the knowledge points and corresponding time periods contained in each teaching video in the recommendation library, and calculate the matching index for each teaching video by substituting them into the formula. :

[0182] ;

[0183] In the formula, The number of reference points contained in the instructional video. The first video in the teaching series The duration of each reference point corresponds to a specific time period. The first video in the teaching series The weighting coefficients of each reference point This refers to the duration of the instructional video.

[0184] Matching Index The weighted knowledge points correspond to the time period percentage in the calculation ( To avoid the brief mention of knowledge points from excessively affecting the recommendation results (e.g., a video that contains reference points but only occupies 10 seconds will be demoted in weight).

[0185] A matching evaluation model for teaching resources and knowledge needs is constructed, with its basic calculation paradigm being a time-weighted integration of each knowledge dimension. All reference knowledge points covered by the teaching videos being evaluated are selected.

[0186] For a single knowledge point, the length of its presentation time in the video is coupled with the corresponding weight value to generate a local matching degree, and global normalization is performed by taking the reciprocal of the total time.

[0187] The local matching degrees of all reference knowledge points are linearly superimposed to form the final matching index. By introducing a time-dimensional weighting mechanism, the depth of knowledge point teaching is accurately represented (e.g., to avoid briefly mentioned knowledge points from excessively affecting the assessment results), while the dynamic injection of weight coefficients achieves adaptive matching with the learner's cognitive state.

[0188] In the specific implementation process, a personalized recommendation strategy based on prioritizing cognitive weaknesses is generated.

[0189] S400: Analyze the changes in the matching index of each instructional video and set a dynamic recommendation list. When a user exits the browsing of instructional videos, freeze the recommendation list and push it to the user. Update behavioral data and attribute parameters based on browsing behavior. Specifically, this includes:

[0190] S401. Calculate the matching index of each teaching video in the recommendation library in real time, and select the top videos in descending order of matching index. Each instructional video is set up with a recommendation list, which is updated in real time based on changes in the matching index.

[0191] S402. When a client exits the tutorial video. When browsing, stop updating and freeze all teaching videos in the recommended list, then push them to the client's browsing interface for display.

[0192] In practice, the list is frozen when exiting the browser to solve the flickering interference problem caused by traditional real-time recommendations and preserve the integrity of the user's intent.

[0193] S403, Update customer browsing history and browsing tutorial videos. Real-time behavioral data is used to update instructional videos after using customer subjects as learning targets. The attribute parameters.

[0194] In the specific implementation process, an interference-free dynamic recommendation closed loop and autonomous algorithm evolution are achieved.

[0195] Example 2: Please refer to Figure 2 The present invention also provides a customer demand intelligent matching management system based on big data, including a data acquisition module, a demand analysis module, an intelligent matching module and a display management module.

[0196] The data acquisition module collects behavioral data of customers browsing instructional videos and analyzes the learning objects and browsing history contained in the instructional videos.

[0197] In the specific implementation process, behavioral data generated by the customer (current learner) when browsing teaching videos is collected, including specific timestamp feedback actions such as submitting question and answer instructions (such as submitting answers), adjusting speed instructions (such as speed setting), and jumping progress instructions (such as dragging the progress bar).

[0198] Simultaneously, the attribute parameters of the instructional videos are parsed, including the video's metadata (such as a list of all learners who have watched the video) and the corresponding browsing history (i.e., the sequence of instructional videos watched by the learners). Acquiring this data relies on direct parsing of the video stream and timestamps, ensuring the real-time nature and completeness of the data collection.

[0199] It provides accurate basic data support, helps the system build a personalized initial data pool by capturing users' real-time interactive behavior, thereby improving the accuracy of subsequent analysis and recommendation modules, avoiding deviations caused by missing data, and ultimately achieving accurate construction of user profiles.

[0200] The requirements analysis module analyzes browsing history and builds a knowledge base, parses the knowledge points contained in teaching videos and constructs a relationship tree diagram, analyzes behavioral data to map knowledge points in the relationship tree diagram and calculates the mastery index.

[0201] In the specific implementation process, teaching videos are selected based on the collected browsing records, and a cognitive library (which stores all videos that users have browsed) and an extended library (which optimizes the number of videos based on the related videos of the learning objects by controlling computing power) are established.

[0202] The video content is then analyzed, and the text is extracted and semantics are integrated using a speech recognition engine and OCR technology. Natural language processing algorithms are then used to segment semantic segments, identify conceptual entities, and construct progressive relationships between knowledge points to form a relational tree diagram.

[0203] At the same time, the mastery index of knowledge points is calculated, and the degree of mastery is quantified by combining variables (such as the accuracy rate of question and answer and the percentage of playback frames).

[0204] It achieves deep modeling of user needs through structured knowledge point dependencies and adaptive parameters (such as positive feedback gain coefficients). and negative feedback suppression coefficient (Dynamic adjustment) accurately maps the user's cognitive weaknesses, improves the system's ability to identify learning progress and difficulties, provides quantifiable evidence for personalized recommendations, and significantly enhances the user's learning efficiency.

[0205] The intelligent matching module is used to build a recommendation library and locate content within a relational tree diagram. Based on this location and the mastery index, it calculates the weight coefficient of each knowledge point and the matching index of each teaching video in the recommendation library.

[0206] In the specific implementation process, based on the video currently being viewed by the user (such as an instructional video) Establish a recommendation library (select relevant videos from the extended library), locate the knowledge point area (reference area) in the relationship tree diagram, and plan the relationship chain from the starting point to the end point.

[0207] The weight of each knowledge point is calculated using a weighting coefficient formula, and then substituted into the matching index formula to evaluate the matching degree of videos in the recommendation library, dynamically adjusting the recommendation priority.

[0208] The algorithm optimizes the recommendation logic, taking into account the coherence of the learning path (such as the dependence of knowledge points between different levels in the relationship chain) and the weight allocation of the mastery index, so as to realize the dynamic generation of personalized video suggestions, effectively improve the real-time performance and relevance of the recommendations, reduce the user's search cost and optimize the learning experience.

[0209] The display management module sets a dynamic recommendation list based on changes in the matching index. When a customer exits the browsing of the teaching videos, the recommendation list is frozen and displayed.

[0210] The display management module calculates the matching index of videos in the recommendation library in real time and dynamically generates a recommendation list (e.g., selecting the top H videos). Updates stop when the user exits the browser to freeze the list, which is then pushed to the user interface. Simultaneously, the module updates the user's browsing history and video attribute parameters (such as the list of learning objects and new behavioral data).

[0211] In the actual implementation process, we ensure the smooth delivery of teaching videos and the responsiveness of the user interface, and improve the user experience by freezing the recommendation list at key moments to avoid interference.

[0212] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0213] Finally, it should be noted that the above descriptions are merely preferred embodiments 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 foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. 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 customer demand intelligent matching management method based on big data, characterized by: The method includes: S100. Collect behavioral data of customer objects when browsing teaching videos, as well as the attribute parameters of the teaching videos; parse the learning objects contained in the attribute parameters, and the browsing records of each learning object; S200: Analyze browsing history to filter teaching videos and establish a cognitive library and an extended library; parse the knowledge points contained in each teaching video and construct a relationship tree diagram; analyze behavioral data to map knowledge points in the relationship tree diagram and calculate the mastery index. S300: Build a recommendation library based on the teaching videos that the customer is currently browsing and locate them in the relationship tree diagram; calculate the weight coefficient of each knowledge point and the matching index of each teaching video in the recommendation library based on the location and the mastery index. S400: Analyze the changes in the matching index of each teaching video and set a dynamic recommendation list. When a customer exits the browsing of teaching videos, freeze the recommendation list and push it to the user. Update behavioral data and attribute parameters based on browsing behavior.

2. The intelligent matching and management method for customer needs based on big data according to claim 1, characterized in that: In S100, the customer refers to the user who is watching the instructional video; the instructional video refers to digital educational resources that use video as a medium to present course content. Behavioral data refers to the feedback actions that users take at different times while watching instructional videos; feedback actions include submitting questions and answers, adjusting speed, and jumping to the next step. The attribute parameters include all learning objects that have watched the corresponding instructional video; the browsing history includes all instructional videos that the learning object has viewed.

3. The intelligent matching and management method for customer needs based on big data according to claim 2, characterized in that: S200 includes: S201. Establish a cognitive database, extract all teaching videos from the customer's browsing history and add them to the cognitive database; count the number of all teaching videos in the cognitive database. And the number of all learners included in each instructional video. ; S202. Establish an extended library, analyze the teaching videos in the browsing history of each learner, and extract these... All instructional videos for each learner have been added to the extended library; among them... ; S203. Analyze the knowledge points and time periods contained in each teaching video in the cognitive library and the extended library, analyze the time sequence of different knowledge points and time periods, build the progressive relationship between knowledge points, and thus construct a relationship tree diagram; S204. Extract behavioral data of customers browsing each teaching video in the cognitive library, analyze the knowledge point time period of each feedback action, calculate the mastery index of each knowledge point, and map them in the relationship tree diagram.

4. The intelligent matching and management method for customer needs based on big data according to claim 3, characterized in that: S203 includes: S2031. Analyze all teaching videos in the cognitive library and the extended library, use a speech recognition engine to convert the speech in the teaching videos into text, and use OCR technology to recognize the images in the teaching videos and extract the text, and merge the texts from the same time period. S2032. Perform semantic segmentation on the fused text, use natural language understanding algorithms to identify semantic conceptual entities, establish logical relationship structure between conceptual entities in combination with contextual relationships, and generate a set of knowledge points. S2033. Synchronously parse the timestamp records of the video stream, take the start and end time nodes of each knowledge point in the video as the time period of its implementation, and output the final time sequence that maps each knowledge point to its corresponding time period. S2034. Arrange all knowledge points under each teaching video according to the time sequence of the time period and construct knowledge chains respectively; analyze the relationship between knowledge points based on all knowledge chains and set progressive relationships, and construct a relationship tree diagram based on the progressive relationships.

5. The intelligent matching and management method for customer needs based on big data according to claim 4, characterized in that: S2034 includes: S2034-1 Obtaining Instructional Videos For all knowledge points in the corresponding knowledge chain, arrange each pair of adjacent knowledge points in chronological order. and Divided into combinations, knowledge points within each combination and They serve as the parent node and child node, respectively. S2034-2. Divide the knowledge chain corresponding to each teaching video into groups, and divide all groups with the same two knowledge points into the same category. Calculate the proportion of the two types of groups in the same category whose parent node and child node have the same knowledge points. S2034-3. Mark classes where the proportion of any combination exceeds a preset threshold, and analyze the knowledge points corresponding to the parent and child nodes of the high-proportion combinations in the marked classes. and Setting up knowledge points arrive The progressive relationship; S2034-4. For unlabeled classes, analyze two knowledge points in the combination and set the peer relationship between these two knowledge points; obtain all knowledge points in all knowledge chains and construct a relationship tree diagram based on the progressive or peer relationship between knowledge points.

6. The intelligent matching and management method for customer needs based on big data according to claim 3, characterized in that: S204 includes: S2041, Extracting Customer Objects and Browsing Tutorial Videos behavioral data Analyze teaching videos Includes time periods for each knowledge point, as well as behavioral data. The time period corresponding to each feedback action in the process; S2042. Calculate the accuracy rate of the questions and answers for each knowledge point. Percentage of unplayed frames Percentage of high-speed playback time Percentage of frames played per session Question and answer error rate Percentage of frames already played Percentage of low-speed playback time and the percentage of frames played multiple times ; S2043, Preset curve steepness factor Positive feedback gain coefficient Negative feedback suppression coefficient and the sharpness of perception of difficulties Tutorial video on substituting formulas for calculation Mastery coefficient of each knowledge point : ； in, and These are the positive feedback factor and the negative feedback factor, respectively, and their calculation formulas are as follows: ； ； In the formula, ; S2044. Similarly, calculate the mastery coefficient for each knowledge point under each teaching video in the cognitive base; calculate the average mastery coefficient of the same knowledge point across all teaching videos in the cognitive base as the mastery index. ; S2045. Map all the mastery indices of all knowledge points to a relational tree diagram.

7. The intelligent matching and management method for customer needs based on big data according to claim 6, characterized in that: S2042 includes: S2042-1. Analyze the knowledge points in the time period corresponding to the submission of the question and answer instruction, and take the percentage of correct answers and the percentage of incorrect answers as the correct answer rate and the incorrect answer rate of the corresponding knowledge points, respectively. S2042-2, Analyze the progress position and adjustment value corresponding to the speed adjustment command, as shown in the teaching video. The timeline is divided into intervals with different playback speeds, and the video playback speed remains constant within the same interval; S2042-3, Preset Normal Playback Speed Playback speed greater than As a high-speed range, the playback speed is less than... As a low-speed range; S2042-4. Divide the total duration of the high-speed interval and the total duration of the low-speed interval in each time period by the duration of the time period to obtain the percentage of high-speed playback time and the percentage of low-speed playback time for the corresponding knowledge points. S2042-5, Analysis of Teaching Videos Includes image frames; analyzes jump progress instructions in the teaching video. The image frames corresponding to the jump-out and jump-in points, as well as whether each image frame is played and the number of times it is played; S2042-6. Count the total number of image frames for each knowledge point within the time period. Divide the number of unplayed image frames and the number of played image frames within the time period by the total number of image frames to obtain the percentage of unplayed frames and the percentage of played frames for the corresponding knowledge point. S2042-7. Divide the number of single-play image frames and the number of multiple-play image frames within the time period by the total number of image frames to obtain the percentage of single-play frames and the percentage of multiple-play frames for the corresponding knowledge points.

8. The intelligent matching and management method for customer needs based on big data according to claim 3, characterized in that: The S300 includes: S301. Obtain the tutorial videos currently being viewed in the customer's browsing history. And mark all the knowledge points it contains; build a recommendation library, and filter out all teaching videos in the extended library that contain at least one marked knowledge point and put them into the recommendation library; S302, Analyzing Teaching Videos The current playback progress is in a specific time period. The corresponding knowledge point in this time period is marked as a positioning point. The positioning point is mapped in the relationship tree diagram, and the area where the positioning point and all its subordinate knowledge points are located is divided as a reference area. S303. Within the reference area, different relationship chains are planned, starting from the positioning point and ending at the knowledge point without subordinates. All knowledge points in each relationship chain are interconnected according to the hierarchical relationship, and the knowledge points in different relationship chains are not completely the same. S304. Take all knowledge points within the reference area, excluding the positioning points, as reference points, and count the number of knowledge points containing reference points. Number of all relationship chains Obtain reference points Mastery Index Calculate the reference point Weighting coefficients : ； In the formula, To include reference points The The number of all reference points in the relationship chain. For the first Reference point in the direction from the start to the end of the relationship chain Calculate the ordinal number; calculate the weight coefficient of each other reference point in the reference area respectively; S305. Analyze the knowledge points and corresponding time periods contained in each teaching video in the recommendation library, and substitute them into the formula to calculate the matching index of each teaching video. : ； In the formula, The number of reference points contained in the instructional video. The first video in the teaching series The duration of each reference point corresponds to a specific time period. The first video in the teaching series The weighting coefficients of each reference point This refers to the duration of the instructional video.

9. The intelligent matching and management method for customer needs based on big data according to claim 8, characterized in that: The S400 includes: S401. Calculate the matching index of each teaching video in the recommendation library in real time, and select the top videos in descending order of matching index. Each instructional video is set into a recommendation list, and the instructional videos in the recommendation list are updated in real time based on changes in the matching index. S402. When a client exits the tutorial video. When browsing, stop updating and freeze all teaching videos in the recommended list, and push them to the customer's browsing interface for display; S403, Update customer browsing history and browsing tutorial videos. Real-time behavioral data is used to update instructional videos after using customer subjects as learning targets. The attribute parameters.

10. A customer demand intelligent matching management system based on big data, characterized by: The system includes a data acquisition module, a demand analysis module, an intelligent matching module, and a display management module; The data acquisition module collects behavioral data of customers browsing instructional videos and parses the learning objects and browsing history contained in the instructional videos; The requirements analysis module analyzes browsing history and builds a knowledge base, parses the knowledge points contained in teaching videos and constructs a relationship tree diagram, analyzes behavioral data to map knowledge points in the relationship tree diagram and calculates the mastery index; The intelligent matching module is used to build a recommendation library and locate the knowledge points in the relationship tree diagram. Based on the location and the mastery index, the weight coefficient of each knowledge point and the matching index of each teaching video in the recommendation library are calculated. The display management module sets a dynamic recommendation list based on changes in the matching index. When a customer exits the browsing of the teaching videos, the recommendation list is frozen and displayed.

Images