Training management system and device based on intelligent recommendation strategy and storage medium

Abstract

The invention discloses a training management system based on an intelligent recommendation strategy, and the system comprises a data processing module which is used for obtaining employee information and course information, and carrying out the preprocessing operation; the course recommendation module is used for converting the employee information and the course information into feature vectors for clustering, and substituting employee class clusters and course class clusters obtained after clustering into a deep neural network model for deep learning so as to obtain recommended courses of the employee class clusters and employees suitable for the course class clusters; the examination supervision module is used for performing an examination after the employees learn the recommended curriculums, and automatically judging cheating and automatically handing over the examination paper when the examinees switch to other programs by using a screen switching prevention technology; and the system updating module is used for updating the system in real time according to the change of the data in the system, regularly evaluating and adjusting a recommendation algorithm and establishing a user feedback mechanism. According to the invention, personalized recommendation can be carried out according to information such as working posts and working years of employees.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recommendation technology, and in particular to a training management system, equipment and storage medium based on an intelligent recommendation strategy. Background Art

[0002] In the current rapidly developing digital age, training management systems have become an important tool for enterprises, educational institutions and other organizations to enhance the capabilities of employees or students. Such systems usually integrate functions such as course management, learning progress tracking, online testing and evaluation, aiming to provide an efficient and flexible learning environment. On the technical level, modern training management systems are mostly built on cloud computing architecture, using big data analysis to personalize learning content recommendations, and optimizing user experience through artificial intelligence technologies such as natural language processing (NLP) and machine learning algorithms. In addition, with the popularization of mobile Internet, support for cross-platform access (including but not limited to PC and mobile APP) has also become a major feature of such systems, which not only greatly enriches the training format, but also significantly improves the training effect. Therefore, training management systems can be widely used in daily use and are suitable for training needs in different fields such as online learning platforms and mental health management.

[0003] For example, in the prior art, a method for intelligently recommending courses on an online learning platform with publication number CN115222564A performs popularity rating on different substitute teachers, and can recommend different teachers and different teaching materials according to the different learning situations of different students, thereby improving the overall learning effect and enabling different students to achieve academic improvement; or, for example, in the prior art, a personalized intelligent recommendation system and method for mental health courses with publication number CN114969545A provides users with similar content recommendation services through keywords.

[0004] However, in actual use, the technologies adopted by the above-mentioned existing technologies either have weak method targeting and scientificity, or have low recommendation accuracy and poor user experience. Different users have very different needs for training resources and cannot meet personalized learning needs.

[0005] To this end, this application specifically proposes a training management system based on intelligent recommendation strategy to solve the above technical problems. Summary of the invention

[0006] The main purpose of the present invention is to provide a training management system based on intelligent recommendation strategy, which improves training efficiency, ensures training quality, and enhances student satisfaction through information technology. Through online course training, examination competitions, test questions, questionnaires and training exchanges, it can achieve full tracking management of student learning status and comprehensive grasp of employee learning and training needs, and finally create a full-process online training and examination one-stop service platform to solve the technical problems raised in the background technology.

[0007] The present invention adopts the following technical solutions to solve the above technical problems:

[0008] A training management system based on intelligent recommendation strategy includes a data processing module, a course recommendation module, an examination supervision module, and a system update module, wherein:

[0009] The data processing module is used to access the external human resources system and obtain employee information and course information from the human resources system, and perform pre-processing operations on the employee information and course information;

[0010] The course recommendation module is used to convert the employee information and course information after basic preprocessing into feature vectors for clustering, and substitute the employee clusters and course clusters obtained after clustering into the deep neural network model for deep learning to obtain the recommended courses for the employee clusters and the employees suitable for the course clusters;

[0011] The examination supervision module uses VUE technology to establish an examination system and implements user interaction through custom services based on JSON. It is used to take examinations for related courses after employees have completed the recommended courses. During the operation of the examination system, the examination supervision module is used to log in using face recognition technology. The examination supervision module has built-in anti-screen switching technology, which is used to automatically determine cheating and automatically hand in the examination paper when the examinee switches to other programs;

[0012] The system update module is used to update the system in real time according to the changes in the data within the system, monitor the system status and operation status in real time, regularly back up and maintain data, regularly update software and hardware, regularly evaluate and adjust the recommendation algorithm, and establish a user feedback mechanism.

[0013] Preferably, the information acquisition and preprocessing operations within the data processing module include:

[0014] Employee information: 4A account, name, department, job title, years of work experience, learning progress, learning time, and test results;

[0015] Course information: course name, course duration, course score;

[0016] Preprocessing operation: Perform data cleaning operations to delete information about employees who have left and inappropriate courses.

[0017] Preferably, the specific steps of processing the internal employee information and course information of the course recommendation module include:

[0018] S1. Convert employee information into feature vectors to obtain employee information feature vectors, use cosine similarity to calculate the similarity of employee information feature vectors, cluster the employee information feature vector similarities using the DBSCAN clustering algorithm to obtain employee clusters, wherein the specific calculation formula for cosine similarity calculation is:

[0019] Among them, x represents the employee information feature vector of the first employee, y represents the employee information feature vector of the second employee, and θ represents the angle with the coordinate axis;

[0020] S2. Convert the course information into a feature vector to obtain a course information feature vector, use cosine similarity to calculate the similarity of the course information feature vector, cluster the course information feature vector similarity using the DBSCAN clustering algorithm to obtain a course cluster, wherein the specific calculation formula for cosine similarity calculation is:

[0021] Where w represents the employee information feature vector of the first employee, z represents the employee information feature vector of the second employee, and μ represents the angle with the coordinate axis;

[0022] S3. construct a deep convolutional neural network model, input the employee cluster and course information feature vector into the deep neural network model, and obtain recommended courses for the employee cluster;

[0023] S4. Input the course cluster and employee information feature vector into a deep neural network model to obtain an employee suitable for the course cluster, and recommend the course cluster to the employee.

[0024] Preferably, the specific execution steps of clustering by the DBSCAN clustering algorithm include:

[0025] L1. Establish a data set for initialization, select data from the data set, calculate the three-dimensional features of user information, and preset a threshold a of the number of users who meet the user conditions;

[0026] L2. Start the loop and preset the clustering parameters e and Minpts of this loop, where e represents the radius of the clustering domain and Minpts represents the minimum number of domain points for a given point to become a core object in the clustering domain;

[0027] L3. Randomly select a user from the data. If the user has not been visited, calculate whether there are at least Minpts data points in the neighborhood of the user. If there are more than Minpts data points, create a new cluster to place the user. Repeat the calculation process for other data points in the e-neighborhood of the user to expand the cluster. When it is detected that the neighborhood without data points can contain at least Minpts data points, the expansion of the cluster ends.

[0028] L4. Repeat step L3 until all users in the data set have been visited, select the cluster that best matches the user's characteristics from the obtained clusters, set the number of clustering generations to K, and perform clustering operations for generations;

[0029] L5. When K>a, it means that the selected users have met the needs, and the iteration ends.

[0030] Preferably, the specific operation steps of performing the clustering operation in the L4 step include:

[0031] In the first clustering, the abnormal user area and the normal user area are selected. At this time, the normal user area will be marked as low-risk area 1, indicating that it is the cluster that best meets the normal user behavior characteristics. In the subsequent clustering process, the abnormal user cluster will no longer be selected, but only the cluster that meets the normal user behavior characteristics will be selected;

[0032] In the process of narrowing the data set, the cluster that meets the malicious user behavior is defined as the medium-risk user area. At this time, the medium-risk user area is not marked. The clusters marked as normal users in this clustering and subsequent clustering will be divided into low-risk area 2, low-risk area 3, and finally low-risk area k. The numerical values ​​of the three-dimensional indicators of users in the above areas are different from those of users in low-risk area 1, and as users of normally running virtual machines, they have the probability of entering low-risk area 1 in the end.

[0033] Preferably, the specific operation steps of constructing the deep convolutional neural network model in the step S3 include: sequentially constructing an embedding layer, a multi-head attention layer, a convolution layer, a pooling layer and an output layer of the convolutional neural network, wherein:

[0034] P1. Embedding layer, which is used to convert the employee information feature vector into a digital matrix according to the length of the feature vector, and then output it to the next layer;

[0035] P2. The multi-head attention layer is used to select important features and obtain potential factors from the employee information feature cluster by introducing a multi-head attention mechanism. The multi-head attention mechanism specifically includes: first, the vector information undergoes a linear transformation, and then it is input into the scaled dot product attention, a total of h times, each time is counted as a head, and the parameters between the heads are not shared. Then, the h times of scaled dot product attention results are spliced, and finally a linear transformation is performed to obtain the value as the result of the multi-head attention, that is:

[0036] head i =Attention(QWQ i ,KWK i ,VWV i )

[0037] Among them: QWQ i Represented as query vector Q i After a linear transformation, KWK i Represented as key vector K i After a linear transformation, VWV i Represented as a value vector V i The result of a linear transformation;

[0038] P3. Convolution layer, used to extract adjacent feature vectors from the feature vector matrix using one-dimensional convolution kernels with window sizes of 1, 3, 5, and 7, thereby extracting context features;

[0039] P4. Pooling layer uses k-max-pooling technology to extract the top k values ​​from each pooling block of the convolutional layer as representative features, and processes documents of variable length by constructing a merging operation of fixed-length feature vectors. The representative features extracted are:

[0040]

[0041] in: Indicates that when the parameters W, X and variance are known The probability of observing V under the condition of , N(·) is expressed as a normal distribution, V j is the jth observation, MACNN W (X j ) refers to the convolutional neural network through which the parameters W are passed, The covariance matrix is The noise model of the times unit matrix I;

[0042] P5. The output layer projects the K-dimensional space of the user’s latent factors for the recommendation task and finally generates the latent vector of the document by using conventional nonlinear projection, thereby transforming the high-level features obtained from the previous layer for specific tasks.

[0043] Preferably, the examination system within the examination supervision module uses video technology combining SRS and WebSocket protocols, wherein:

[0044] On the SRS server, WebSocket establishes a connection through the HTTP protocol and then upgrades to the WebSocket protocol;

[0045] When establishing a connection, the client sends an HTTP request to the server, and the request header contains the Upgrade field, whose value is WebSocket;

[0046] After receiving the request, if the server supports the WebSocket protocol, it returns an HTTP response. The response header contains the Upgrade field, whose value is WebSocket. The response header also contains the Sec-WebSocket-Accept field, whose value is obtained by encrypting the Sec.WebSocket-Key field sent by the client and a fixed GU ID string.

[0047] After the client receives the response, if the response header contains the Upgrade field, the value is WebSocket, and the value of the Sec-WebSocket-Accept field is the same as the encrypted result of the Sec-WebSocket-Key field sent by the client, it means that the connection has been successfully established and WebSocket communication can begin.

[0048] Preferably, the anti-screen cutting technology inside the test supervision module is used for anti-screen cutting detection in the web program, specifically including:

[0049] The page currently being used by the user is defined as a focus page, where the focus page covers the common page;

[0050] Check whether the focus page status is maintained during the exam. If the focus page status is lost during the process, it is regarded as the candidate has performed a screen switching operation, triggering the screen switching cheating rules.

[0051] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the internal algorithm steps of the above system.

[0052] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the internal algorithm steps of the above system.

[0053] It can be seen from the above technical solution that the present invention provides a training management system based on intelligent recommendation strategy. Compared with the prior art, the present invention has the following advantages:

[0054] 1. The present invention sets the DBSCAN clustering algorithm and the deep convolutional neural network model in the course recommendation module, which can accurately identify the characteristics of employees and courses and perform effective clustering. It can make personalized recommendations based on information such as employees' job positions and years of work experience. It can recommend suitable courses to similar employee clusters, and can also recommend course clusters to suitable employees. Therefore, the comprehensive dual recommendation can improve the accuracy of the recommendation, thereby improving the pertinence and satisfaction of employee learning.

[0055] 2. The present invention sets a multi-head attention mechanism in the course recommendation module, which can play a role in obtaining potential factors from employee information feature clusters, thereby achieving the effect of improving the accuracy of the recommendation algorithm, and finally better meeting the learning needs of different employees.

[0056] 3. The present invention adopts Vue technology, JSON interaction, SRS and video technology combined with WebSocket protocol in the examination supervision module, which can improve the system response speed and user interaction experience, thereby achieving real-time and efficient examination supervision effect, and finally ensuring the fairness and security of the examination process.

[0057] 4. The present invention can prevent cheating by setting up face recognition login and anti-screen cutting detection technology in the examination supervision module, thereby achieving the effect of maintaining the fairness of the examination, and finally ensuring the authenticity and credibility of the examination results.

[0058] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become easy to understand through the following description. Of course, it is not necessary to achieve all of the advantages described above simultaneously for any product implementing the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0060] Figure 1It is an overall schematic diagram of the steps of using the system of the present invention;

[0061] Figure 2 It is an overall schematic diagram of the system module framework of the present invention;

[0062] Figure 3 This is an overall schematic diagram of the internal system operation steps of the course recommendation module of the present invention. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. In the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0064] In the embodiment, see Figures 1 to 3 .

[0065] like Figure 1 and Figure 2 As shown, a training management system based on an intelligent recommendation strategy proposed in an embodiment of the present invention includes a data processing module, a course recommendation module, an examination supervision module, and a system update module, wherein:

[0066] (1) A data processing module, which is used to access the external human resources system and obtain employee information and course information from the human resources system, and perform pre-processing operations on the employee information and course information;

[0067] Employee information includes: 4A account, name, department, job title, years of work experience, learning progress, learning time, and test results;

[0068] The course information includes: course name, course duration, and course score;

[0069] The pre-processing operations include: performing data cleaning operations, such as deleting information of employees who have left the company and inappropriate course information;

[0070] In summary, the data processing module can perform data cleaning by setting up pre-processing functions, delete information about resigned employees and unsuitable courses, improve data quality and accuracy, and thus ensure that the data foundation for subsequent recommendation and supervision modules is more reliable;

[0071] (2) Course recommendation module, which is used to convert the employee information and course information after basic preprocessing into feature vectors for clustering, and substitute the employee clusters and course clusters obtained after clustering into the deep neural network model for deep learning to obtain the recommended courses for the employee clusters and the employees suitable for the course clusters;

[0072] The specific steps for processing employee information and course information in the course recommendation module are as follows: Figure 3 As shown, including:

[0073] S1. Convert employee information into feature vectors to obtain employee information feature vectors, use cosine similarity to calculate the similarity of employee information feature vectors, cluster the employee information feature vector similarity using the DBSCAN clustering algorithm to obtain employee clusters, where the specific calculation formula for cosine similarity calculation is:

[0074] Among them, x represents the employee information feature vector of the first employee, y represents the employee information feature vector of the second employee, and θ represents the angle with the coordinate axis;

[0075] S2. Convert the course information into a feature vector to obtain the course information feature vector, use cosine similarity to calculate the similarity of the course information feature vector, cluster the similarity of the course information feature vector using the DBSCAN clustering algorithm to obtain the course cluster, where the specific calculation formula for cosine similarity calculation is:

[0076] Where w represents the employee information feature vector of the first employee, z represents the employee information feature vector of the second employee, and μ represents the angle with the coordinate axis;

[0077] It should be specifically explained here that the specific execution steps of clustering by the DBSCAN clustering algorithm in the above steps S1 and S2 include:

[0078] L1. Establish a data set for initialization, select data from the data set, calculate the three-dimensional features of each user's information, determine the number of users who roughly meet the user conditions by observing the data, and preset a threshold a for the number of users who meet the user conditions;

[0079] L2. Start the loop and preset the clustering parameters e and Minpts of this loop, where e represents the radius of the clustering domain and Minpts represents the minimum number of domain points for a given point to become a core object in the clustering domain;

[0080] L3. Randomly select a user from the data. If the user has not been visited, calculate whether there are at least Minpts data points in the neighborhood of the user. If there are more than Minpts data points, create a new cluster to place the user. Repeat the calculation process for other data points in the e-neighborhood of the user to expand the cluster. When it is detected that the neighborhood without data points can contain at least Minpts data points, the expansion of the cluster ends.

[0081] L4. Repeat step L3 until all users in the data set have been visited, select the cluster that best matches the user's characteristics from the obtained clusters, set the number of clustering generations to K, and perform clustering operations for generations;

[0082] In particular, the specific steps of performing clustering operations at this time include:

[0083] In the first clustering, the abnormal user area and the normal user area are selected. The normal user area at this time will be marked as low-risk area 1, indicating that it is the cluster that best meets the normal user behavior characteristics. In the subsequent clustering process, the abnormal user cluster will no longer be selected, but only the cluster that meets the normal user behavior characteristics will be selected. This is because, due to the reduction in the scope of the data set, the cluster that meets the malicious user behavior at this time cannot be directly defined as an abnormal user, but should be defined as a medium-risk user area. The medium-risk user area does not need to be marked. The clusters marked as normal users in this clustering and subsequent clustering will be divided into low-risk area 2, low-risk area 3, and even low-risk area K (K is the number of clustering generations). Although these users still have the behavior characteristics of normal users, the numerical values ​​of their three-dimensional indicators may be different from those of users in low-risk area 1, but these users are still users who operate virtual machines normally and may eventually enter low-risk area 1;

[0084] L5. When K>a, it means that the selected users have met the needs, and the iteration ends;

[0085] S3. Build a deep convolutional neural network model, input the employee cluster and course information feature vector into the deep neural network model, and obtain the recommended courses for the employee cluster;

[0086] It can also be added here that the specific steps of building a deep convolutional neural network model include: sequentially building the embedding layer, multi-head attention layer, convolution layer, pooling layer and output layer of the convolutional neural network, where:

[0087] P1. Embedding layer, which is used to convert the employee information feature vector into a digital matrix according to the length of the feature vector, and then output it to the next layer. For example, if there is a feature vector, the embedding vector in the employee information feature is represented by a matrix. Then, the feature vector is further trained through the optimization process;

[0088] P2. The multi-head attention layer is used to select important features. By introducing a multi-head attention mechanism between the embedding layer and the convolution layer of the convolutional neural network, potential factors can be obtained from the employee information feature cluster. The multi-head attention mechanism specifically includes: first, the vector information is subjected to a linear transformation, and then it is input into the scaled dot product attention, a total of h times, each time is counted as a head, and the parameters between the heads are not shared. Then, the h times of scaled dot product attention results are spliced, and finally a linear transformation is performed to obtain the value as the result of the multi-head attention. The difference of the multi-head attention at this time is that h calculations are performed instead of just once, that is:

[0089] head i =Attention(QWQ i ,KWK i ,VWV i )

[0090] Among them: QWQ i Represented as query vector Q i After a linear transformation, KWK i Represented as key vector K i After a linear transformation, VWV i Represented as a value vector V i The result of a linear transformation;

[0091] P3. Convolution layer, used to extract adjacent feature vectors from the feature vector matrix using one-dimensional convolution kernels with window sizes of 1, 3, 5, and 7, thereby extracting context features;

[0092] P4. Pooling layer uses k-max-pooling technology to extract the top k values ​​from each pooling block of the convolutional layer as representative features, and processes documents of variable length by constructing a merging operation of fixed-length feature vectors. The representative features extracted are:

[0093]

[0094] in: Indicates that when the parameters W, X and variance are known The probability of observing V under the condition of , N(·) is expressed as a normal distribution, V j is the jth observation, MACNN W (X j ) refers to the convolutional neural network through which the parameters W are passed, The covariance matrix is The noise model of the times unit matrix I;

[0095] P5. Output layer, which projects the K-dimensional space of the user’s latent factors for the recommendation task, and finally generates the latent vector of the document by using conventional nonlinear projection, thereby transforming the high-level features obtained from the previous layer for specific tasks;

[0096] Therefore, by setting up a multi-head attention mechanism in the course recommendation module, it is possible to obtain potential factors from employee information feature clusters, thereby improving the accuracy of the recommendation algorithm and better meeting the learning needs of different employees;

[0097] S4. Same as step S3, input the course cluster and employee information feature vector into the deep neural network model, obtain the employee suitable for the course cluster, and recommend the course cluster to the employee;

[0098] In summary, the course recommendation module can accurately identify employee and course characteristics and perform effective clustering by setting feature vector conversion, cosine similarity calculation, DBSCAN clustering algorithm and deep convolutional neural network model. It can make personalized recommendations based on employee job positions, years of work and other information. It can recommend suitable courses to similar employee clusters, and can also recommend course clusters to suitable employees. Therefore, the comprehensive dual recommendation can improve the accuracy of the recommendation, thereby improving the pertinence and satisfaction of employee learning.

[0099] (3) Examination supervision module: The examination system is established using VUE technology and is based on JSON for interaction. In addition, user interaction is achieved through a variety of services and custom services. It is used to allow employees to take examinations for related courses after they have completed the recommended courses. During the operation of the examination system, the examination supervision module is used to log in using face recognition technology. The examination supervision module also has built-in anti-screen switching technology, which is used to automatically determine cheating and automatically submit the examination paper when the examinee switches to other programs;

[0100] Specifically, the examination system inside the examination supervision module uses the video technology combining SRS and WebSocket protocol, in which: on the SRS server, WebSocket establishes a connection through the HTTP protocol and then upgrades to the WebSocket protocol; when establishing a connection, the client sends an HTTP request to the server, and the request header contains an Upgrade field, whose value is WebSocket; after the server receives the request, if it supports the WebSocket protocol, it returns an HTTP response, and the response header contains an Upgrade field, whose value is WebSocket, and the response header also contains a Sec-WebSocket-Accept field, whose value is obtained by encrypting the Sec.WebSocket-Key field sent by the client and a fixed GU ID string; after the client receives the response, if the response header contains an Upgrade field, whose value is WebSocket, and the value of the Sec-WebSocket-Accept field is the same as the encrypted result of the Sec-WebSocket-Key field sent by the client, it means that the connection has been successfully established and WebSocket communication can begin;

[0101] Therefore, the examination supervision module adopts Vue technology, JSON interaction, SRS and WebSocket protocol combined video technology to improve the system response speed and user interaction experience, thereby achieving real-time and efficient examination supervision effect, thereby ensuring the fairness and security of the examination process;

[0102] Specifically, the anti-screen cutting technology inside the examination supervision module is used for anti-screen cutting detection in the web program, including: defining the page currently used by the user as the focus page, and the focus page covers the ordinary page. At this time, the anti-screen cutting detection is for the network examination software to check whether it maintains the "focus page" status; then check whether the focus page status is maintained during the examination. If the network examination software loses the "focus page" status at a certain moment, it is regarded as the candidate performing a screen cutting operation, thereby triggering the screen cutting cheating rules. This is the principle of the anti-screen cutting function for online examinations. For example, in a Windows system, if a user opens multiple folders, the top color of the folder being operated will become darker and pinned to the top, while the tops of other folders that have not been operated will be displayed in light colors. This is the difference between the focus page and the ordinary page. The same is true between programs. The focus page usually covers the ordinary page, which is more convenient for users to use, and the network examination software itself is just a web program;

[0103] Therefore, by setting up face recognition login and anti-screen cutting detection technology in the examination supervision module, it can prevent cheating, thereby maintaining the fairness of the examination and ensuring the authenticity and credibility of the examination results;

[0104] (4) System update module, which is used to update the system in real time according to the changes in the data in the system, monitor the system status and operation status in real time, regularly back up and maintain data, regularly update software and hardware, regularly evaluate and adjust the recommendation algorithm, and establish a user feedback mechanism;

[0105] At this time, the system update module can maintain stable system operation and continuous optimization by setting up real-time data updates, regular backup and maintenance, software and hardware updates, system status monitoring, and user feedback mechanisms, thereby improving the overall system performance and user experience, and ensuring the long-term reliability of the system and user satisfaction.

[0106] In summary, in actual use, the system can carry the training course system built by the enterprise based on job requirements, and record various internal and external training course materials (including attached documents, videos, etc.). The training courses and their materials can be made public online to designated employees as learning materials, and also include the management of training instructors and training channels related to the training courses. In addition, the construction of the training system can link the training content with job competency and quality, making the training work systematic and providing a good user experience.

[0107] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0108] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0109] In another embodiment provided by the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute the internal module algorithm steps of any of the training management systems based on intelligent recommendation strategies in the above embodiments.

[0110] It is understandable that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above method.

[0111] The embodiment of the present application also provides an electronic device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus.

[0112] Memory, used to store computer programs;

[0113] The processor is used to implement the internal module algorithm steps of the training management system based on the intelligent recommendation strategy when executing the program stored in the memory.

[0114] The communication bus mentioned in the above electronic device may be a Peripheral Component Interconnect (PCI) bus or an Extended Industrial Standard Architecture (EISA) bus, etc. The communication bus may be divided into an address bus, a data bus, a control bus, etc.

[0115] The communication interface is used for communication between the above electronic device and other devices.

[0116] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0117] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0118] It should also be noted that electronic devices also include terminal devices, which can also be called terminals, user equipment, mobile stations, mobile terminals, etc. Terminal devices can be mobile phones, smart TVs, wearable devices, tablet computers, computers with wireless transceiver functions, virtual reality terminal devices, augmented reality terminal devices, wireless terminals in industrial control, wireless terminals in unmanned driving, wireless terminals in remote surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, etc. The embodiments of this application do not limit the specific technology and specific device form used by the terminal devices.

[0119] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server, data center, etc. that contains one or more available media integrated. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium, or a semiconductor medium, etc.

[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0121] In addition, it should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components in a certain specific posture. If the specific posture changes, the directional indication will also change accordingly.

[0122] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the meaning of "and / or" appearing in the full text includes three parallel schemes. Taking "A and / or B" as an example, it includes scheme A, or scheme B, or schemes that A and B meet at the same time. In addition, in the embodiments of the present invention, "multiple" refers to more than two. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

Claims

1. A training management system based on intelligent recommendation strategy, characterized in that: It includes data processing module, course recommendation module, examination supervision module and system update module, among which: The data processing module is used to access the external human resources system and obtain employee information and course information from the human resources system, and perform pre-processing operations on the employee information and course information; The course recommendation module is used to convert the employee information and course information after basic preprocessing into feature vectors for clustering, and substitute the employee clusters and course clusters obtained after clustering into the deep neural network model for deep learning to obtain the recommended courses for the employee clusters and the employees suitable for the course clusters; The examination supervision module uses VUE technology to establish an examination system and implements user interaction through custom services based on JSON. It is used to take examinations for related courses after employees have completed the recommended courses. During the operation of the examination system, the examination supervision module is used to log in using face recognition technology. The examination supervision module has built-in anti-screen switching technology, which is used to automatically determine cheating and automatically hand in the examination paper when the examinee switches to other programs; The system update module is used to update the system in real time according to the changes in the data within the system, monitor the system status and operation status in real time, regularly back up and maintain data, regularly update software and hardware, regularly evaluate and adjust the recommendation algorithm, and establish a user feedback mechanism.

2. The training management system based on intelligent recommendation strategy as claimed in claim 1, characterized in that: The data processing module internally acquires information and performs preprocessing operations including: Employee information: 4A account, name, department, job title, years of work experience, learning progress, learning time, and test results; Course information: course name, course duration, course score; Preprocessing operation: Perform data cleaning operations to delete information about employees who have left and inappropriate courses.

3. The training management system based on intelligent recommendation strategy as claimed in claim 1, characterized in that: The specific steps of processing the internal employee information and course information of the course recommendation module include: S1. Convert employee information into feature vectors to obtain employee information feature vectors, use cosine similarity to calculate the similarity of employee information feature vectors, cluster the employee information feature vector similarities using the DBSCAN clustering algorithm to obtain employee clusters, wherein the specific calculation formula for cosine similarity calculation is: Among them, x represents the employee information feature vector of the first employee, y represents the employee information feature vector of the second employee, and θ represents the angle with the coordinate axis; S2. Convert the course information into a feature vector to obtain a course information feature vector, use cosine similarity to calculate the similarity of the course information feature vector, cluster the course information feature vector similarity using the DBSCAN clustering algorithm to obtain a course cluster, wherein the specific calculation formula for cosine similarity calculation is: Where w represents the employee information feature vector of the first employee, z represents the employee information feature vector of the second employee, and μ represents the angle with the coordinate axis; S3. construct a deep convolutional neural network model, input the employee cluster and course information feature vector into the deep neural network model, and obtain recommended courses for the employee cluster; S4. Input the course cluster and employee information feature vector into a deep neural network model to obtain an employee suitable for the course cluster, and recommend the course cluster to the employee.

4. The training management system based on intelligent recommendation strategy as claimed in claim 3, characterized in that: The specific execution steps of clustering by the DBSCAN clustering algorithm include: L1. Establish a data set for initialization, select data from the data set, calculate the three-dimensional features of user information, and preset a threshold a of the number of users who meet the user conditions; L2. Start the loop and preset the clustering parameters e and Minpts for this loop, where e represents the radius of the clustering domain and Minpts represents the minimum number of domain points for a given point to become a core object in the clustering domain; L3. Randomly extract a user from the data. If the user has not been visited, calculate whether there are at least Minpts data points in the neighborhood of the user. If there are more than Minpts data points, create a new cluster to put the user in. Repeat the calculation process for other data points in the e-neighborhood of the user to expand the cluster. When it is detected that the neighborhood without data points can contain at least Minpts data points, the expansion of the cluster ends. L4. Repeat step L3 until all users in the data set have been visited, select the cluster that best matches the user's characteristics from the obtained clusters, set the number of clustering generations to K, and perform clustering operations for generations; L5. When K>a, it means that the selected users have met the needs, and the iteration ends.

5. The training management system based on intelligent recommendation strategy as claimed in claim 4, characterized in that: The specific operation steps of performing the clustering operation in the L4 step include: In the first clustering, the abnormal user area and the normal user area are selected. At this time, the normal user area will be marked as low-risk area 1, indicating that it is the cluster that best meets the normal user behavior characteristics. In the subsequent clustering process, the abnormal user cluster will no longer be selected, but only the cluster that meets the normal user behavior characteristics will be selected; In the process of narrowing the data set, the cluster that meets the malicious user behavior is defined as the medium-risk user area. At this time, the medium-risk user area is not marked. The clusters marked as normal users in this clustering and subsequent clustering will be divided into low-risk area 2, low-risk area 3, and finally low-risk area k. The numerical values ​​of the three-dimensional indicators of users in the above areas are different from those of users in low-risk area 1, and as users of normally running virtual machines, they have the probability of entering low-risk area 1 in the end.

6. The training management system based on intelligent recommendation strategy as claimed in claim 3, characterized in that: The specific operation steps of constructing the deep convolutional neural network model in step S3 include: sequentially constructing an embedding layer, a multi-head attention layer, a convolution layer, a pooling layer, and an output layer of the convolutional neural network, wherein: P1. Embedding layer, which is used to convert the employee information feature vector into a digital matrix according to the length of the feature vector, and then output it to the next layer; P2. The multi-head attention layer is used to select important features and obtain potential factors from the employee information feature cluster by introducing a multi-head attention mechanism. The multi-head attention mechanism specifically includes: first, the vector information undergoes a linear transformation, and then it is input into the scaled dot product attention, a total of h times, each time is counted as a head, and the parameters between the heads are not shared. Then, the h times of scaled dot product attention results are spliced, and finally a linear transformation is performed to obtain the value as the result of the multi-head attention, that is: head i =Attention(QWQ i ,KWK i ,VWV i ) Among them: QWQ i Represented as query vector Q i After a linear transformation, KWK i Represented as key vector K i After a linear transformation, VWV i Represented as a value vector V i The result of a linear transformation; P3. Convolution layer, used to extract adjacent feature vectors from the feature vector matrix using one-dimensional convolution kernels with window sizes of 1, 3, 5, and 7, thereby extracting context features; P4. Pooling layer uses k-max-pooling technology to extract the top k values ​​from each pooling block of the convolutional layer as representative features, and processes documents of variable length by constructing a merging operation of fixed-length feature vectors, where the representative features extracted are: in: Indicates that when the parameters W, X and variance are known The probability of observing V under the condition of , N(·) is expressed as a normal distribution, V j is the jth observation, MACNN W (X j ) refers to the convolutional neural network through which the parameters W are passed, The covariance matrix is The noise model of the times unit matrix I; P5. The output layer projects the K-dimensional space of the user’s latent factors for the recommendation task and finally generates the latent vector of the document by using conventional nonlinear projection, thereby transforming the high-level features obtained from the previous layer for specific tasks.

7. The training management system based on intelligent recommendation strategy according to claim 1, characterized in that: The examination system within the examination supervision module uses video technology that combines SRS and WebSocket protocols, where: On the SRS server, WebSocket establishes a connection through the HTTP protocol and then upgrades to the WebSocket protocol; When establishing a connection, the client sends an HTTP request to the server, and the request header contains the Upgrade field, whose value is WebSocket; After receiving the request, if the server supports the WebSocket protocol, it returns an HTTP response. The response header contains the Upgrade field, whose value is WebSocket. The response header also contains the Sec-WebSocket-Accept field, whose value is obtained by encrypting the Sec.WebSocket-Key field sent by the client and a fixed GUID string. After the client receives the response, if the response header contains the Upgrade field, the value is WebSocket, and the value of the Sec-WebSocket-Accept field is the same as the encrypted result of the Sec-WebSocket-Key field sent by the client, it means that the connection has been successfully established and WebSocket communication can begin.

8. The training management system based on intelligent recommendation strategy as claimed in claim 1, characterized in that: The anti-screen cutting technology inside the test supervision module is used for anti-screen cutting detection in the web program, specifically including: The page currently being used by the user is defined as a focus page, where the focus page covers the common page; Check whether the focus page status is maintained during the exam. If the focus page status is lost during the process, it is regarded as the candidate has performed a screen switching operation, triggering the screen switching cheating rules.

9. A computer-readable storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a processor, the processor executes the internal algorithm steps of the system according to any one of claims 1 to 8.

10. A computer device, characterized in that: The system comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the internal algorithm steps of the system according to any one of claims 1 to 8.

Images