A management method and storage medium for an educational institution management system.

Abstract

This application relates to the field of educational institution management, and more particularly to a management method and storage medium for an educational institution management system. The method includes: collecting parents' location data and pre-filled information through promotional links; the management system performing primary and secondary matching based on the location data and pre-filled information, where primary matching is based on location data and secondary matching is based on pre-filled information; and adjusting the weight allocation of primary and secondary matching based on historical recommendation samples and payment conversion success rates using a recommendation strategy optimization algorithm, generating a list of recommended educational institutions and pushing it to parents' users. Parents then visit the recommended educational institutions to complete check-in and payment. This achieves closed-loop management from promotional links to recommended institutions to payment conversion, improving the management efficiency of the educational institution management system and increasing the payment conversion success rate through the matching mechanism and optimization algorithm.

Description

Technical Field

[0001] This application relates to the field of educational institution management, and more particularly to a management method and storage medium for an educational institution management system. Background Technology

[0002] In the education industry, most operations rely on a combination of social media, paper documents, and back-end systems to manage parents and students. This approach suffers from data silos, leading to information chaos, inefficiency, and significant manpower costs.

[0003] Existing management systems for educational institutions can achieve end-to-end management of promotion, payment, and course management. However, these systems are often migrated from e-commerce or recommendation systems and are not well-suited for educational institutions. They cannot support the complete management cycle of educational institutions and fail to meet their needs for management efficiency and payment conversion success rates. In conclusion, there is no existing management system or related management method suitable for educational institutions. Summary of the Invention

[0004] To address the lack of suitable management systems and methods for educational institutions in the existing technology, this application provides a management method and storage medium for an educational institution management system. The method provided by this application includes the following steps: The system collects parents' location data and pre-filled information through promotional links. The system performs a first match and a second match based on the location data and the pre-filled information. The first match is based on the location data, and the second match is based on the pre-filled information. The recommendation strategy optimization algorithm adjusts the weight allocation of the first and second matching based on historical recommendation samples and payment conversion success rates, generates a list of recommended educational institutions, and pushes it to the parent's user terminal. Parents go to the educational institutions to complete check-in and payment according to the list of recommended educational institutions. The recommendation strategy optimization algorithm treats parents who go to the educational institution to check in and pay as a first-class sample, parents who go to the educational institution but do not pay as a second-class sample, and parents who do not go to the educational institution as a third-class sample, and optimizes the weight allocation strategy of the first and second matching based on the first-class, second-class, and third-class samples.

[0005] Specifically, during the first matching, the coordinates of the educational institutions registered in the management system are obtained, and the parent's location data is matched with the coordinates. During the matching, a Hidden Markov Model (HMM) is used to correct the parent's location data. The observation probability of the HMM is calculated based on the Euclidean distance between the trajectory of the parent's location data and the topological nodes in the road network. The transition probability of the HMM is obtained based on path connectivity training and is used as a reliability index of the first matching to be input into the recommendation strategy optimization algorithm. When the confidence level is lower than the threshold, the weight allocation ratio of the first matching is reduced and the weight ratio of the second matching is increased.

[0006] Specifically, when using a hidden Markov model to modify the road network, the transition probability satisfies: , in, s i , s j For adjacent road segment nodes, d ij This represents the actual distance between road sections. m The average spacing of the road network is obtained through training on historical trajectory data. During training, two training sets are distinguished: one for peak school dismissal times and another for off-peak times. The appropriate training set is selected based on the time when the parent's user accesses the promotional link. m value.

[0007] Specifically, the pre-filled information includes the student's age and course preferences. When performing the secondary matching, the course settings and age range in the educational institution's registration information are obtained, and the student's age is matched with the age range, and the course preferences are matched with the course settings.

[0008] Specifically, when performing the secondary matching, a first deviation value is calculated between the student's age and the age range, a second deviation value is calculated between the course preference and the course settings, and the service rating of the educational institution is obtained. The first deviation value, the second deviation value, and the service rating are weighted and a matching degree is calculated.

[0009] Specifically, the recommendation strategy optimization algorithm adjusts the weight ratio of the first matching and the second matching by minimizing the weighted cross-entropy loss function of positive samples and negative samples. The weighted cross-entropy loss function is a function that reflects the distribution of positive samples and negative samples. The positive samples are the set of the first type of samples, and the negative samples are the set of the second type of samples and the third type of samples.

[0010] Specifically, when the recommendation strategy optimization algorithm optimizes the weight allocation strategy for the first and second matching, it assigns different weight values ​​in the weighted cross-entropy loss function to the first-class samples, the second-class samples, and the third-class samples, and adds a time decay coefficient to the historical recommendation samples according to the time order.

[0011] Specifically, the management method further includes: When parents pay fees, their signatures are entered. The image of the signature is enhanced using a pix2pix model, which includes a self-attention module and a folding operation. The self-attention module is used to improve the edge clarity of calligraphy characters, and the folding operation is used to simulate the ink diffusion effect when writing on paper.

[0012] Specifically, the management method further includes: After parents complete the payment, the system generates a student file and stores it in association with the parent's signature when making the payment, for use in student management and renewal management.

[0013] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described above.

[0014] This application has the following technical advantages: It achieves closed-loop management from promotional links to recommended institutions and then to payment conversion, improving the management efficiency of educational institutions' management systems. Through matching mechanisms and optimization algorithms, it forms the algorithmic support for the educational institution recommendation list, thereby increasing the success rate of payment conversion. Attached Figure Description

[0015] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. Several embodiments of this application are illustrated in the drawings by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts.

[0016] Figure 1 This is a flowchart of a management method for an educational institution management system, as described in an embodiment of this application. Detailed Implementation

[0017] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0018] In the wave of digitalization, educational institutions often build a series of management systems to improve their management level. These management systems involve multiple business processes, such as accessing WeChat promotional links, recommending educational institutions, and managing students after payment, achieving full-process management. Among these business processes, recommending educational institutions is the most relevant to the profitability of educational institutions, as it directly affects parents' willingness to visit educational institutions and the matching degree between parents and educational institutions, thus affecting the success rate of conversion to payment. Based on this, this embodiment aims to propose a management method for educational institutions' management systems, which can not only meet the management needs of the entire process, but also effectively improve the conversion rate of parents to payment in the process of recommending educational institutions. Figure 1 As shown, the management method in this embodiment includes the following steps: The system collects parents' location data and pre-filled information through promotional links. The system performs a first match and a second match based on the location data and pre-filled information. The first match is based on the location data, and the second match is based on the pre-filled information. The recommendation strategy optimization algorithm adjusts the weight allocation of primary and secondary matching based on historical recommendation samples and payment conversion success rates, generating a list of recommended educational institutions and pushing it to parents' users. Parents then go to the recommended educational institutions to complete check-in and payment. The recommendation strategy optimization algorithm treats parents who go to the educational institution to check in and pay as a first-class sample, parents who go to the educational institution but do not pay as a second-class sample, and parents who do not go to the educational institution as a third-class sample, and optimizes the weight allocation strategy of primary and secondary matching based on the first-class, second-class, and third-class samples.

[0019] Specifically, in this embodiment, when parents click on promotional links placed by educational institutions through social platforms such as WeChat, the system first calls the GPS module of the mobile terminal to obtain real-time location coordinates. Simultaneously, it guides parents to fill in core pre-filled information such as the student's age, desired course type, and available time period. This data is uploaded to the cloud server via an encrypted transmission protocol, triggering the map matching engine's computation process. Since the parents' movement trajectory is continuous and constrained, while GPS / base station positioning data contains random noise, this embodiment uses a Hidden Markov Model (HMM) to express the parents' state transition probabilities and observation probabilities. The parent's location data and coordinates are matched, and the HMM is used to correct the parent's location data during matching. The observation probability of the HMM is calculated based on the Euclidean distance between the parent's location data trajectory and the topological nodes in the road network, while the transition probability is obtained through path connectivity training. Specifically, the HMM model models the urban road network topology as a state space, with each road segment node corresponding to a hidden state. The observation sequence represents the coordinates of the parent's movement trajectory. The transition probability matrix is ​​obtained through training with historical traffic flow data, and the transition probabilities satisfy the following: , in, s i , s j For adjacent road segment nodes, d ij This represents the actual distance between road sections. m The average spacing of the road network was obtained through training on historical trajectory data. During training, two training sets were distinguished: one for peak school dismissal times and another for off-peak times. The appropriate training set was selected based on the time when parents accessed the promotional links on their devices. m The value, typically, is obtained from the training set during the peak school dismissal time on weekdays, from 3:00 PM to 6:00 PM. m The value is less than that obtained from the training set during off-peak hours. m Value. Meanwhile, m The confidence level can also be used as a reliability indicator for the first match and input into the recommendation strategy optimization algorithm. When the confidence level is lower than the threshold, the weight allocation ratio of the first match is reduced, and the weight allocation ratio of the second match is increased. This effectively solves the problem of institution matching deviation caused by the complexity of urban road networks and location signal drift in educational scenarios. Furthermore, it distinguishes between historical trajectory data during peak and off-peak hours after school and forms a dynamic compensation mechanism. This significantly improves the matching accuracy of the educational institution recommendation list during peak hours in densely populated school districts. The secondary conversion rate of unpaid visitors increases due to the improved path correction accuracy, breaking through the application bottleneck of traditional recommendation algorithms in complex urban educational scenarios.

[0020] After the first matching is completed, the system enters the second matching stage. The core objective here is to deeply correlate the student's age and course preferences in the parents' pre-filled information with the educational institution's course offerings, age range, and service ratings in the registration information. For example, if a parent pre-fills in a 6-year-old student's age and a course preference of "English learning for beginners," the system will prioritize matching with educational institutions offering "early childhood English" courses with an age range of 5-7 years old. To achieve accurate matching, the system calculates a first deviation value between the student's age and the institution's age range. The specific formula is as follows: in, A 机构中值 This represents the midpoint of the institution's age-suitable range. A 机构范围The age range is considered. Similarly, course preference matching extracts keywords (such as "early childhood education" and "art") using natural language processing technology and calculates the cosine similarity with the labels in the institution's course offerings to obtain the second deviation value Δ2. Service rating directly uses the institution's historical rating data (such as parent reviews and course completion rates) to calculate its third deviation value Δ3 compared to the industry benchmark rating. Finally, the system generates a comprehensive matching score based on a dynamic weight allocation mechanism. Initial weights are set to w 1 = 0.3 w 2 = 0.25 w =3=0.2, the remaining weight is reserved for extended parameters.

[0021] In the weight allocation stage, the recommendation strategy optimization algorithm uses a double forgetting factor mechanism to dynamically adjust the weight ratio of geographic matching and demand matching. Initially, the weight for geographic matching is set to 0.6, and the weight for demand matching is 0.4, updated every 24 hours based on the latest conversion data. For continuous optimization of the recommendation strategy, the system retrains the model every 500 new samples. During training, a cost-sensitive learning method is used for the three types of samples: the weight of successfully paid samples (Type 1) is set to 1.5, the weight of unpaid samples (Type 2) is 0.8, and the weight of unvisited samples (Type 3) is 0.3. The loss function is modified as follows: , in, y ∈{0, 1} represents the true label (positive or negative sample). p It is the model's predicted probability. w + and w − These are the weighting coefficients for positive and negative samples, respectively.

[0022] When parents visit educational institutions based on the recommended list, the system automatically checks them in via Bluetooth beacon interaction with the parent's mobile app. Upon successful check-in, a payment guidance process is triggered, allowing parents to complete their handwritten signature on the electronic agreement. The signature processing module employs an improved pix2pix generative adversarial network, innovatively incorporating a folding operation subnetwork into the generator network. This subnetwork decomposes the input strokes into two feature channels: horizontal and vertical strokes, simulating the pressure variations during real writing through deformable convolutional layers. The discriminator network uses a multi-scale PatchGAN structure, calculating adversarial loss in a 256×256 pixel local region, effectively mitigating the legal risks associated with distorted handwriting in traditional electronic signatures.

[0023] After payment is completed, the system automatically generates a student profile and stores the signature image encrypted using a hash algorithm, linked to the student's information. In subsequent course management, the teacher's app can retrieve historical attendance records, course progress, and other data from the student profile in real time.

[0024] In summary, this embodiment constructs a complete and efficient management system for educational institutions by integrating high-precision map matching, dynamic weighted recommendation strategies, and intelligent signature processing technology. This system not only solves the data silo problem of traditional management methods but also significantly improves payment conversion rates through algorithm-driven accurate recommendations, providing reliable technical support for the digital transformation of the education industry.

[0025] Obviously, the embodiments described above are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0026] It should be understood that when the terms "first," "second," etc., are used in the claims, description, and drawings of this application, they are only used to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" used in the description and claims of this application indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

Claims

1. A management method for an educational institution management system, characterized in that, Includes the following steps: The system collects pre-filled information from parents through promotional links and uses the GPS module of the mobile terminal to obtain the parents' real-time location data. The system performs a first matching and a second matching based on the location data and the pre-filled information. The first matching is based on the location data, and the second matching is based on the pre-filled information. The recommendation strategy optimization algorithm adjusts the weight allocation of the first and second matching based on historical recommendation samples and payment conversion success rates, generates a list of recommended educational institutions, and pushes it to the parent user's terminal. Parents go to the educational institutions to complete check-in and payment according to the list of recommended educational institutions. The recommendation strategy optimization algorithm treats parents who go to the educational institution to check in and pay as a first-class sample, parents who go to the educational institution but do not pay as a second-class sample, and parents who do not go to the educational institution as a third-class sample, and optimizes the weight allocation strategy of the first and second matching based on the first-class, second-class, and third-class samples. During the first matching, the coordinates of the educational institutions registered in the management system are obtained, and the location data of the parents is matched with the coordinates. During the matching, a Hidden Markov Model (HMM) is used to correct the location data of the parents. The observation probability of the HMM is calculated based on the Euclidean distance between the trajectory of the parents' location data and the topological nodes in the road network. The transition probability of the HMM is obtained based on path connectivity training and is used as the reliability index of the first matching to be input into the recommendation strategy optimization algorithm. When the confidence is lower than the threshold, the weight allocation ratio of the first matching is reduced and the weight ratio of the second matching is increased. When using a Hidden Markov Model to modify the road network, the transition probabilities satisfy: ， in, s i , s j For adjacent road segment nodes, d ij This represents the actual distance between road sections. μ The average spacing of the road network is obtained through training on historical trajectory data. During training, two training sets are distinguished: one for peak school dismissal times and another for off-peak times. The appropriate training set is selected based on the time when the parent's user accesses the promotional link. μ The μ value is used as a reliability indicator for a single match and input into the recommendation strategy optimization algorithm. The peak school dismissal period is from 15:00 to 18:00 on weekdays, and the μ value obtained from the training set during the peak school dismissal period is less than the μ value obtained from the training set during the off-peak period.

2. The management method according to claim 1, characterized in that, The pre-filled information includes the student's age and course preferences. When performing the secondary matching, the course settings and age range in the educational institution's registration information are obtained, and the student's age is matched with the age range, and the course preferences are matched with the course settings.

3. The management method according to claim 2, characterized in that, When performing the secondary matching, a first deviation value is calculated between the student's age and the age range, a second deviation value is calculated between the course preference and the course settings, and the service rating of the educational institution is obtained. The first deviation value, the second deviation value, and the service rating are weighted and a matching degree is calculated.

4. The management method according to claim 1, characterized in that, The recommendation strategy optimization algorithm adjusts the weight ratio of the first matching and the second matching by minimizing the weighted cross-entropy loss function of positive and negative samples. The weighted cross-entropy loss function is a function that reflects the distribution of positive and negative samples. The positive samples are the set of the first type of samples, and the negative samples are the set of the second type of samples and the third type of samples.

5. The management method according to claim 4, characterized in that, When optimizing the weight allocation strategy for the first and second matching, the recommendation strategy optimization algorithm assigns different weight values ​​in the weighted cross-entropy loss function to the first-class samples, the second-class samples, and the third-class samples, and adds a time decay coefficient to the historical recommendation samples according to the time order.

6. The management method according to claim 1, characterized in that, The management method also includes: When parents pay fees, their signatures are entered. The image of the signature is enhanced using a pix2pix model, which includes a self-attention module and a folding operation. The self-attention module is used to improve the edge clarity of calligraphy characters, and the folding operation is used to simulate the ink diffusion effect when writing on paper.

7. The management method according to claim 1, characterized in that, The management method also includes: After parents complete the payment, the system generates a student file and stores it in association with the parent's signature when making the payment, for use in student management and renewal management.

8. A computer-readable storage medium storing a computer program, characterized in that, When the program is executed by the processor, it implements the steps of the management method according to any one of claims 1-7.

Images