AI-based computational rules and systems for key indicators in preschool education

Abstract

This invention discloses a method and system for invoking computational rules based on key indicators of preschool education, relating to the field of preschool education technology. It involves constructing a cloud-based big data center to collect multi-source data from preschool education, processing this data to obtain correlated data clusters, and building a preschool education AI model including a judgment model, a child development assessment and diagnosis model, a personalized gamified learning content recommendation model, and a home-school collaborative interaction support model. The judgment model analyzes the input to the preschool education AI model to determine which type of model to use for key indicator output. Simultaneously, a dynamic data pipeline is constructed to connect the cloud-based big data center and the preschool education AI, and a detection window is set to protect data security. This invention achieves more intelligent and precise preschool education intervention through the cloud-based big data center, the preschool education AI model, and the dynamic data pipeline.

Description

Technical Field

[0001] This invention relates to the field of preschool education technology, specifically to a method and system for AI-based computational rules based on key indicators of preschool education. Background Technology

[0002] Currently, the assessment and intervention of children's development in the field of preschool education mainly relies on manual observation, recording, and analysis, which is inefficient and highly subjective. Traditional data collection methods are scattered and incomplete, making it difficult to integrate data from various aspects such as individual children, teaching environments, families and communities, and physical environments. This results in a lack of in-depth understanding of children's development. In addition, existing data analysis methods are usually relatively simple, making it difficult to uncover the complex relationships and patterns hidden in the data, and failing to provide effective decision support for personalized education. More importantly, data security and privacy protection face serious challenges during data transmission and use. Traditional data transmission methods are vulnerable to attacks and tampering, lack effective access control and security protection mechanisms, and pose a risk of data leakage.

[0003] Therefore, it is of great significance to construct an AI-based algorithm and system that uses key indicators of preschool education. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for AI-based call operation rules based on key indicators of preschool education, in order to address the shortcomings in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an AI-based computational rule method for calling operations based on key indicators of preschool education, comprising: constructing a cloud-based big data center and collecting multi-source data on preschool education;

[0006] Build an AI model for preschool education and define the rules for its invocation;

[0007] A dynamic data pipeline is constructed, through which the cloud-based big data center and the preschool education AI model interact, including:

[0008] Data ports are set up on both the cloud big data center and the preschool education AI model, and the data ports of the cloud big data center and the preschool education AI model are connected through dynamic data pipelines;

[0009] A comparison unit is set up in the data port of the cloud big data center. The comparison unit is used to receive the call data sent by the preschool education AI model, and to perform semantic analysis on key indicators to retrieve similar mapping indicators and obtain the corresponding data keys.

[0010] The data key is used to access the indicator atlas and call the corresponding data cluster;

[0011] The requested data set is transmitted as response data to the preschool education AI model via a dynamic data pipeline. A detection window is set on the dynamic data pipeline to confirm the safety of the data set.

[0012] In a preferred embodiment, the step of constructing a cloud-based big data center and collecting multi-source data on preschool education is as follows:

[0013] Collect multi-source data on preschool education and store it in a cloud-based big data center;

[0014] Multi-source data includes individual child data, teaching environment data, family and community data, and physical environment data;

[0015] Feature extraction is performed on multi-source data to obtain mapping indicators, correlation analysis is performed on the mapping indicators to obtain the correlation matrix, and data clusters are constructed based on the correlation matrix.

[0016] In a preferred embodiment, the steps of extracting features from multi-source data to obtain mapping indicators, performing correlation analysis on the mapping indicators to obtain an association matrix, and constructing data clusters based on the association matrix are as follows:

[0017] The analysis of variance and principal component analysis were used to extract features from multi-source data to obtain mapping indicators;

[0018] Define the set of mapping indicators Calculate the correlation coefficient between each pair of mapping indicators, and construct an association matrix based on the correlation coefficient. :

[0019] ,in, Represents the correlation matrix. This represents the correlation coefficient between the mapping indicators;

[0020] Preset coefficient thresholds, construct an indicator atlas by mapping indicators whose correlation coefficients exceed the coefficient thresholds, and connect the multi-source data pointed to by the indicator atlases to generate data clusters.

[0021] The data cluster includes multi-source data pointed to by the indicator atlas and the directed edges connecting them, which are composed of the values ​​of the correlation coefficients.

[0022] In a preferred embodiment, the steps of constructing the preschool education AI model and defining the calling rules are as follows:

[0023] The construction of AI models for preschool education includes judgment models, child development assessment and diagnosis models, personalized gamified learning content recommendation models, and home-school co-education interaction support models.

[0024] Among them, preset key indicator tables are set for the child development assessment and diagnosis model, the personalized gamified learning content recommendation model, and the home-school co-education interaction support model;

[0025] The judgment model processes the input to the preschool education AI model to generate key indicators and call requests.

[0026] In a preferred embodiment, the step of processing the input of the preschool education AI model based on the judgment model to generate key indicators and call requests is as follows:

[0027] The judgment model receives input from the preschool education AI model and extracts keywords through semantic segmentation algorithms;

[0028] The semantic similarity algorithm is used to determine the similarity between keywords and key indicators in the key indicator table, and the similarity values ​​are recorded and sorted to obtain a similarity ranking table.

[0029] A preset similarity threshold is set, and keywords that exceed the similarity threshold are selected as similarity indicators based on the similarity ranking table. The key indicators pointed to by the similarity indicators are recorded.

[0030] Based on the key indicator comparison table pointed to by similar indicators, a call request is sent to the model to which the key indicator table belongs, and the call request and key indicators are used as call data.

[0031] In a preferred embodiment, the step of transmitting the called data group as response data to the preschool education AI model through a dynamic data pipeline, and setting a detection window on the dynamic data pipeline, is as follows:

[0032] The response data is stored in the comparison unit, the comparison unit is used to encapsulate the response data and apply an audit watermark layer, and the audit watermark layer is used to wrap the encapsulated response data to obtain the encapsulated response data.

[0033] The encapsulated response data is transmitted through a dynamic data pipeline, and multiple detection windows are opened on the dynamic data pipeline.

[0034] When the encapsulated response data passes through the detection window, a unique ID scratch of the detection window will be engraved on the audit watermark layer, and all detection window ID scratches engraved on the audit watermark layer will be identified.

[0035] If scratches appear on the identified audit watermark layer that do not belong to the detection window of the dynamic data pipeline, the detection window will destroy the encapsulated response data, generate a detection log including scratch information, and send a retransmission request to the cloud big data center.

[0036] In a preferred embodiment, an AI-based invocation and computation rule system based on key indicators of preschool education is constructed, including:

[0037] Cloud server module: Constructs a cloud-based big data center to collect multi-source data on preschool education;

[0038] Early Childhood Education AI Module: Construct early childhood education AI models and define invocation rules;

[0039] Data security module: Constructs a dynamic data pipeline through which the cloud-based big data center and the preschool education AI model interact, including:

[0040] Data ports are set up on both the cloud big data center and the preschool education AI model, and the data ports of the cloud big data center and the preschool education AI model are connected through dynamic data pipelines;

[0041] A comparison unit is set up in the data port of the cloud big data center. The comparison unit is used to receive the call data sent by the preschool education AI model, and to perform semantic analysis on key indicators to retrieve similar mapping indicators and obtain the corresponding data keys.

[0042] The data key is used to access the indicator atlas and call the corresponding data cluster;

[0043] The requested data set is transmitted as response data to the preschool education AI model via a dynamic data pipeline. A detection window is set on the dynamic data pipeline to confirm the safety of the data set.

[0044] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0045] 1. This invention, by constructing a cloud-based big data center, can collect and integrate preschool education data from multiple sources. These multi-source data include individual child data such as health and developmental assessments; teaching environment data such as teacher-student interaction and teaching resources; family and community data such as family background and community support; and physical environment data such as classroom facilities and outdoor activity areas. Integrating this diverse data breaks down the traditional data silos in education, providing a possibility for a comprehensive understanding of children's development. The solution utilizes ANOVA and principal component analysis to extract features from the multi-source data, obtaining mapping indicators. Correlation analysis is then used to construct an association matrix and data clusters. This deep feature extraction and association analysis can reveal patterns and rules hidden in complex data. For example, it can analyze the relationship between family economic status, parents' education level, and children's language development. Through this in-depth analysis, educators can more accurately identify key factors affecting children's development, providing a scientific basis for developing personalized education plans. Furthermore, the construction of data clusters also facilitates rapid retrieval and access to relevant data, improving data utilization efficiency.

[0046] 2. This invention, by constructing a preschool education AI model and determining the model's calling rule mechanism, can significantly improve the accuracy and efficiency of calling preschool education AI models. The model extracts keywords from the preschool education AI input through a semantic segmentation algorithm and uses a semantic similarity algorithm to match these keywords with indicators in a preset key indicator table. By setting a similarity threshold, the system can accurately select the key indicators most relevant to the input and send a calling request to the corresponding AI model accordingly. This method avoids indiscriminate triggering of all AI models, thereby greatly reducing unnecessary consumption of computing resources. The key indicator extraction and similarity judgment process ensures that only the AI ​​model most suitable for processing the current input will be called.

[0047] 3. This invention constructs a dynamic data pipeline connecting a cloud-based big data center and a preschool education AI model to achieve secure and reliable data transmission. During data transmission, a comparison unit is set up in the data port of the cloud-based big data center to perform semantic analysis on key indicators, retrieve similar mapping indicators, and obtain the corresponding data keys, thereby achieving access control over data. Furthermore, to ensure data security during transmission, this solution also employs data encapsulation, audit watermarking layers, and detection windows. Specifically, the response data is stored in the comparison unit, which encapsulates the response data and assigns an audit watermark layer. Multiple detection windows are opened on the dynamic data pipeline. By marking the unique ID of the detection window on the audit watermark layer, it is possible to identify whether the data has been tampered with or leaked. If the detection window identifies a scratch that does not belong to the detection window of its dynamic data pipeline, the encapsulated response data is immediately destroyed, and a detection log and retransmission request are generated. This multi-layered security protection mechanism effectively ensures data security and privacy, providing strong support for the promotion and application of preschool education AI. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0049] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0051] Example 1, please refer to Figure 1 As shown in this embodiment, the AI-based computational rule method for invoking key indicators of preschool education includes:

[0052] S1. Construct a cloud-based big data center to collect multi-source data on preschool education;

[0053] S2. Construct a preschool education AI model and define the calling rules;

[0054] S3. Construct a dynamic data pipeline, through which the cloud-based big data center and the preschool education AI model interact.

[0055] As described in steps S1-S3 above, currently, the assessment and intervention of children's development in the field of preschool education mainly relies on manual observation, recording, and analysis, which is inefficient and highly subjective. Traditional data collection methods are scattered and incomplete, making it difficult to integrate data from various aspects such as individual children, teaching environments, families and communities, and physical environments. This results in an insufficient understanding of children's development. In addition, existing data analysis methods are usually relatively simple, making it difficult to uncover the complex relationships and patterns hidden in the data, and failing to provide effective decision support for personalized education. More importantly, data security and privacy protection face serious challenges during data transmission and use. Traditional data transmission methods are vulnerable to attacks and tampering, lack effective access control and security protection mechanisms, and pose a risk of data leakage.

[0056] This invention, by constructing a cloud-based big data center, can collect and integrate preschool education data from multiple sources. These multi-source data include individual child data such as health and developmental assessments; teaching environment data such as teacher-student interaction and teaching resources; family and community data such as family background and community support; and physical environment data such as classroom facilities and outdoor activity areas. Integrating this diverse data breaks down the traditional data silos in education, making it possible to comprehensively understand children's development. The solution utilizes analysis of variance and principal component analysis to extract features from the multi-source data, obtaining mapping indicators. Correlation analysis is then used to construct association matrices and data clusters. This deep feature extraction and association analysis can reveal patterns and rules hidden in complex data. For example, it can analyze the relationship between family economic status, parents' education level, and children's language development. Through this in-depth analysis, educators can more accurately identify key factors affecting children's development, providing a scientific basis for developing personalized education plans. Furthermore, the construction of data clusters also facilitates rapid retrieval and access to relevant data, improving data utilization efficiency.

[0057] By constructing an AI model for preschool education and determining the model's calling rules mechanism, the accuracy and efficiency of calling the preschool education AI model can be significantly improved. The model extracts keywords from the preschool education AI input through semantic segmentation algorithms and uses semantic similarity algorithms to match these keywords with indicators in a preset key indicator table. By setting a similarity threshold, the system can accurately select the key indicators most relevant to the input and send a calling request to the corresponding AI model accordingly. This method avoids indiscriminate triggering of all AI models, thereby greatly reducing unnecessary consumption of computing resources. The key indicator extraction and similarity judgment process ensures that only the AI ​​model most suitable for processing the current input will be called.

[0058] By constructing a dynamic data pipeline connecting a cloud-based big data center and an early childhood education AI model, secure and reliable data transmission is achieved. During data transmission, a comparison unit is set up in the data port of the cloud-based big data center to perform semantic analysis on key indicators to retrieve similar mapping indicators and obtain the corresponding data keys, thereby achieving access control over data. Furthermore, to ensure data security during transmission, the solution also employs technologies such as data encapsulation, audit watermarking layers, and detection windows. Specifically, the response data is stored in the comparison unit, which encapsulates the response data and assigns an audit watermark layer. Multiple detection windows are opened on the dynamic data pipeline. By scratching a unique ID mark on the audit watermark layer for each detection window, it is possible to identify whether the data has been tampered with or leaked. If a detection window identifies a mark that does not belong to the detection window of its dynamic data pipeline, the encapsulated response data is immediately destroyed, and a detection log and retransmission request are generated. This multi-layered security protection mechanism effectively ensures data security and privacy, providing strong support for the promotion and application of early childhood education AI.

[0059] In one embodiment, step S1 of constructing a cloud-based big data center and collecting multi-source data from preschool education includes:

[0060] S11. Collect multi-source data on preschool education and save it to the cloud big data center;

[0061] S12. Multi-source data includes individual child data, teaching environment data, family and community data, and physical environment data;

[0062] S13. Extract features from multi-source data to obtain mapping indicators, perform correlation analysis on the mapping indicators to obtain the correlation matrix, and construct data clusters based on the correlation matrix;

[0063] As described in steps S11-S13 above, the technical solution for constructing a cloud-based big data center to collect multi-source data in preschool education first involves data collection. This requires collecting data from multiple sources, including individual child data, teaching environment data, family and community data, and physical environment data, and securely storing it in the cloud-based big data center. Individual child data includes the child's age, gender, developmental assessment results, and health records. Teaching environment data includes class size, teacher-student ratio, teacher qualifications, teaching methods, and curriculum. Family and community data includes family income, parents' education level, community resources, and parental involvement. Physical environment data includes classroom area, lighting, noise levels, and safety facilities. Subsequently, through feature extraction, key features are extracted from these multi-source data and mapped into quantifiable indicators. Then, correlation analysis is performed to determine the relationships between these indicators, constructing a correlation matrix. Based on this, data clusters are formed, organizing related data together for subsequent analysis and application.

[0064] In one embodiment, step S13, which involves extracting features from multi-source data to obtain mapping indicators, performing correlation analysis on the mapping indicators to obtain a correlation matrix, and constructing a data cluster based on the correlation matrix, includes:

[0065] S131. Using analysis of variance and principal component analysis, feature extraction is performed on multi-source data to obtain mapping indicators;

[0066] S132, Define the set of mapping indicators Calculate the correlation coefficient between each pair of mapping indicators, and construct an association matrix based on the correlation coefficient. :

[0067]

[0068] S133, where, Represents the correlation matrix. This represents the correlation coefficient between the mapping indicators;

[0069] S134. Preset coefficient threshold, construct an indicator atlas from the mapping indicators whose correlation coefficient exceeds the coefficient threshold, and connect the multi-source data pointed to by the indicator atlas to generate a data cluster.

[0070] S135. The data cluster includes multi-source data pointed to by the indicator atlas and the directed edges connecting them, which are composed of the values ​​of the correlation coefficients.

[0071] As described in steps S131-S135 above, the aim is to extract key information from multi-source data and organize it into data clusters that are easy to understand and analyze. This process first uses analysis of variance (ANOVA) and principal component analysis (PCA) to perform dimensionality reduction and feature selection on the multi-source data. ANOVA is used to identify whether there are significant differences between different categories of data and to extract features with discriminative power. PCA transforms high-dimensional data into a few principal components through linear transformation, preserving the main information of the data and thus obtaining more representative mapping indicators. Subsequently, a set of mapping indicators is defined, and the Pearson correlation coefficient between each pair of indicators in the set is calculated. This coefficient reflects the degree of linear correlation between the indicators, and its value ranges from [value range missing]. From -1 to 1, a correlation matrix is ​​constructed based on these correlation coefficients. To filter out indicators with strong correlations, a coefficient threshold is preset. Mapping indicators whose absolute correlation coefficients exceed this threshold are included in the indicator atlas. The indicator atlas can be understood as a collection of key indicators. Finally, based on the indicator atlas, the multi-source data pointed to by the indicators in the atlas are connected by directed edges to form data clusters. Each data cluster contains the multi-source data pointed to by the indicator atlas, as well as the directed edges connecting these data. The weights of the directed edges are composed of the correlation coefficient values, representing the strength and direction of the correlation between the data. This data cluster structure can clearly show the inherent connections between multi-source data, facilitating subsequent data analysis and mining.

[0072] In one embodiment, step S2, which involves constructing a preschool education AI model and defining invocation rules, includes:

[0073] S21. Constructing AI models for preschool education includes judgment models, child development assessment and diagnosis models, personalized gamified learning content recommendation models, and home-school co-education interaction support models;

[0074] S22, which includes a preset key indicator table for the child development assessment and diagnosis model, the personalized gamified learning content recommendation model, and the home-school co-education interaction support model;

[0075] S23. Based on the judgment model, process the input of the preschool education AI model to generate key indicators and call requests.

[0076] As described in steps S21-S23 above, the aim is to construct four key AI models to support a comprehensive preschool education experience. This process includes building a judgment model, a child development assessment and diagnosis model, a personalized gamified learning content recommendation model, and a home-school collaboration and interaction support model. This model serves as the "command center" of the entire AI system, responsible for receiving input, analyzing the context, and deciding which sub-models to invoke. The child development assessment and diagnosis model comprehensively analyzes individual child data, including developmental assessments, learning behaviors, and health status, combined with data from the teaching environment, family, and community, to assess children's cognitive, language, social-emotional, and motor development levels and diagnose potential developmental problems. Its implementation method uses Bayesian networks to establish a dependency model between different areas of child development for comprehensive assessment and diagnosis. The personalized gamified learning content recommendation model recommends personalized gamified learning content based on children's developmental assessment results, interests, learning behavior data, and combined with the teaching environment and family resources, stimulating children's learning interest and promoting their all-round development. Its implementation method uses reinforcement learning networks. By interacting with children and continuously optimizing recommendation strategies to improve learning outcomes, the home-school co-education interaction support model provides parents with personalized parenting advice and interactive support based on children's developmental assessment results, learning progress, family environment, and community resources. This promotes home-school collaboration and jointly fosters children's healthy growth. The model utilizes natural language generation to generate personalized parenting advice and activity recommendations based on each child's individual circumstances. Simultaneously, three models are used as sub-models, and a key indicator table is constructed for each sub-model. This table reflects the sub-model's function and serves as a judgment standard for the model. For example, the key indicator table for the child development assessment and diagnosis model may include cognitive development level, language development level, and motor development level; the key indicator table for the personalized gamified learning content recommendation model may include interest areas, gamified element preferences, and learning style preferences; and the key indicator table for the home-school co-education interaction support model includes parenting styles, quality of parent-child interaction, and family environment support. Finally, the judgment model processes the preschool education AI input and selects the most suitable sub-model based on the comparison of the key indicator tables.

[0077] In one embodiment, step S23, which processes the input of the preschool education AI model based on the judgment model to generate key indicators and invocation requests, includes:

[0078] S231. The judgment model receives the input from the preschool education AI model and extracts keywords through semantic segmentation algorithm;

[0079] S232. Use a semantic similarity algorithm to determine the similarity between keywords and key indicators in the key indicator table, record the similarity values ​​and sort them to obtain a similarity ranking table.

[0080] S233. Preset a similarity threshold, select keywords that exceed the similarity threshold as similarity indicators according to the similarity ranking table, and record the key indicators pointed to by the similarity indicators.

[0081] S234. Based on the key indicators pointed to by similar indicators, compare the key indicator table and send a call request to the model to which the key indicator table belongs, and use the call request and key indicators as call data.

[0082] As described in steps S231-S234 above, the model needs to be able to receive input from multiple sources, including text data such as teachers' observation records, parents' feedback, and children's voice input, and structured data such as children's age, gender, and learning records. A semantic segmentation algorithm is used to identify keywords in the input text. Transformer models, such as BERT and GPT, can be used for semantic segmentation to better understand the semantic information of the text. Next, a semantic similarity algorithm is used to determine the similarity between the keywords and the key indicators in the key indicator table. This algorithm measures the degree of similarity between the extracted keywords and the key indicators in the table. Commonly used semantic similarity algorithms include Word2Vec, GloVe, and BERT. The key indicator table contains predefined key indicators and their descriptions, reflecting the functions and focus of the preschool education AI model. For each extracted keyword, its semantic similarity with all key indicators in the key indicator table is calculated. The calculated similarity values ​​are then sorted to generate a similarity ranking table, which includes the keywords and key indicators. The system establishes the correspondence between indicators and their similarity values. A preset similarity threshold is used to filter keywords that are sufficiently similar to the key indicators. The similarity threshold needs to be adjusted according to the actual situation. If the threshold is too high, it may fail to identify effective similar indicators, while if the threshold is too low, it may introduce noise. A dynamic threshold can be considered, for example, adjusting the threshold based on factors such as keyword frequency and the weight of the key indicators. Based on the similarity ranking table, keywords with similarity values ​​exceeding the threshold are selected as similar indicators, and the key indicator pointed to by each similar indicator is recorded. A call request is sent to the model to which the key indicator table belongs based on the key indicator pointed to by the similar indicator. Furthermore, if a similar indicator points to multiple key indicators, and these key indicators belong to different models, a certain strategy is needed to select which model to call. Priority can be set for different models, prioritizing models with higher priority. Alternatively, a vote can be held based on the number of similar indicators pointing to different models, selecting the model with the most votes. Multiple models can be called simultaneously, and their outputs are merged. The call request and key indicators are then passed as call data to the corresponding sub-model.

[0083] In one embodiment, step S3, which involves constructing a dynamic data pipeline connecting a cloud-based big data center and a preschool education AI model, includes:

[0084] S31. Data ports are set up on both the cloud big data center and the preschool education AI model. The data ports of the cloud big data center and the preschool education AI model are connected through dynamic data pipelines.

[0085] S32. Set up a comparison unit in the data port of the cloud big data center. The comparison unit is used to receive the call data sent by the preschool education AI model, including the call application and key indicators, and to perform semantic analysis on the key indicators to retrieve similar mapping indicators and obtain the corresponding data keys.

[0086] S33. The data key is used to access the indicator atlas and call the corresponding data group;

[0087] S34. The called data group is transmitted to the preschool education AI model as response data through the dynamic data pipeline. A detection window is set on the dynamic data pipeline to confirm the safety of the data group.

[0088] As described in steps S31-S34 above, a dynamic data pipeline is constructed to connect the cloud big data center and the preschool education AI model. First, data ports are set up on both the cloud big data center and the AI ​​model, and connected through the dynamic data pipeline. A comparison unit is set up on the data port of the big data center. This unit receives the call data from the AI ​​model, including the call request and key indicators, and performs semantic analysis on the key indicators to retrieve similar mapping indicators. Then, the corresponding data key is obtained. The data key is a credential for accessing a specific data group. The data key contains access permission information to ensure that only authorized users can access the data. The data key is associated with the mapping indicators to ensure that only models related to the key indicators can access the corresponding data. The data key is used to access the indicator atlas, verify the validity of the key, and call the corresponding data group. Finally, the called data group is transmitted as response data to the preschool education AI model through the dynamic data pipeline. At the same time, a detection window is set up on the data pipeline to ensure the security of the data group during transmission. This process realizes the AI ​​model's dynamic and secure access to the big data center data.

[0089] In one embodiment, step S34, which involves transmitting the invoked data as response data to the preschool education AI model via a dynamic data pipeline and setting a detection window on the dynamic data pipeline, includes:

[0090] S341. Store the response data in the comparison unit, encapsulate the response data using the comparison unit and assign it to the audit watermark layer, and wrap the encapsulated response data with the audit watermark layer to obtain the encapsulated response data.

[0091] S342. Transmit the encapsulated response data through a dynamic data pipeline, and open multiple detection windows on the dynamic data pipeline;

[0092] S343. When the encapsulated response data passes through the detection window, a unique ID scratch of the detection window will be engraved on the audit watermark layer, and all detection window ID scratches engraved on the audit watermark layer will be identified.

[0093] S344. If scratches appear on the identified audit watermark layer that do not belong to the detection window of the dynamic data pipeline, the detection window will destroy the encapsulated reply data, generate a detection log including scratch information, and send a retransmission request to the cloud big data center.

[0094] As described in steps S341-S344 above, to ensure secure data transmission, the data packets retrieved from the cloud big data center are first stored as response data in the comparison unit. The comparison unit encapsulates this response data and assigns an audit watermark layer, thus forming encapsulated response data. The principle of the audit watermark layer is to embed a specific digital signal into the response data using a specific watermarking algorithm. The receiver can verify the integrity and source of the data by detecting this signal. Subsequently, the encapsulated response data is transmitted through a dynamic data pipeline. Multiple detection windows are pre-opened on this pipeline, and each detection window is assigned a unique ID. When the encapsulated response data passes through any detection window, the window will engrave its unique ID scratch on the audit watermark layer and simultaneously identify all existing detections on the audit watermark layer. The window ID scratch detection mechanism uses a watermarking algorithm to detect audit watermark layers. Each window encrypts its unique ID using an encryption algorithm and then embeds it into the audit watermark layer as an ID scratch. If the detection window identifies an ID scratch on the audit watermark layer that does not belong to this dynamic data pipeline detection window, it immediately determines that the data may have been tampered with or leaked. The detection window will immediately destroy the encapsulated response data and generate a detailed detection log, including all identified scratch information. Simultaneously, the detection window will send a retransmission request to the cloud big data center to ensure that the AI ​​model receives complete and secure data, thereby guaranteeing the reliability and security of the entire system. This multi-detection mechanism effectively prevents data from being illegally tampered with or stolen during transmission, enhancing the overall security protection capabilities of the system.

[0095] In one embodiment, a system for constructing an AI-invoking computation rule method based on key indicators of preschool education includes:

[0096] Cloud server module: Constructs a cloud-based big data center to collect multi-source data on preschool education;

[0097] Early Childhood Education AI Module: Construct early childhood education AI models and define invocation rules;

[0098] Data security module: Constructs dynamic data pipelines through which cloud-based big data centers and preschool education AI models interact.

[0099] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An AI invocation operation rule method based on pre-school education key indicators, characterized in that: a cloud big data center is constructed, and multi-source data of pre-school education is collected; a pre-school education AI model is constructed, and an invocation rule is defined; a dynamic data pipeline is constructed, and the cloud big data center and the pre-school education AI model interact through the dynamic data pipeline, including: data ports are set on the cloud big data center and the pre-school education AI model, and the data ports of the cloud big data center and the pre-school education AI model are connected through the dynamic data pipeline; a comparison unit is set in the data port of the cloud big data center, which is used to receive the invocation data sent by the pre-school education AI model, and to perform semantic analysis on the key indicators to retrieve similar mapping indicators to obtain corresponding data keys; the data keys are used to access the indicator album and invoke the corresponding data group; the invoked data group is transmitted to the pre-school education AI model as reply data through the dynamic data pipeline, and a detection window is set on the dynamic data pipeline, which is used to confirm the safety of the data group.

2. The pre-school education key indicator-based AI call operation rule method according to claim 1, characterized in that: The step of constructing the cloud big data center and collecting multi-source data of pre-school education is: collecting multi-source data of pre-school education and saving it to the cloud big data center; the multi-source data includes child individual data, teaching environment data, family and community data, and physical environment data; feature extraction is performed on the multi-source data to obtain mapping indicators, correlation analysis is performed on the mapping indicators to obtain a correlation matrix, and a data group is constructed based on the correlation matrix.

3. The pre-school education key indicator-based AI call operation rule method according to claim 2, characterized in that: The step of performing feature extraction on the multi-source data to obtain mapping indicators, performing correlation analysis on the mapping indicators to obtain a correlation matrix, and constructing a data group based on the correlation matrix is: variance analysis and principal component analysis are used to perform feature extraction on the multi-source data to obtain mapping indicators; Defining a mapping indicator set , calculating a correlation coefficient between each two mapping indicators, and constructing a correlation matrix according to the correlation coefficient : ; wherein, denotes a correlation matrix, denotes a correlation coefficient between the mapping indicators; a preset coefficient threshold is set, and the mapping indicators with a correlation coefficient exceeding the coefficient threshold are used to form an indicator album, and the multi-source data pointed to by the indicator album is connected to generate a data group; the data group includes the multi-source data pointed to by the indicator album and the directed edges connecting them, and the directed edges are composed of the values of the correlation coefficients.

4. The pre-school education key indicator-based AI call operation rule method according to claim 1, characterized in that: The step of constructing the pre-school education AI model and defining the invocation rule is: the pre-school education AI model includes a judgment model, a child development assessment and diagnosis model, a personalized game-based learning content recommendation model, and a home co-education interactive support model; a preset key indicator table is set for the child development assessment and diagnosis model, the personalized game-based learning content recommendation model, and the home co-education interactive support model; the judgment model processes the input of the pre-school education AI model to generate key indicators and invocation applications.

5. The pre-school education key indicator-based AI call operation rule method according to claim 4, characterized in that: The step of processing the input of the pre-school education AI model based on the judgment model to generate key indicators and invocation applications is: the judgment model receives the input of the pre-school education AI model, and extracts keywords through semantic segmentation algorithm; the semantic similarity algorithm is used to judge the similarity between the keywords and the key indicators in the key indicator table, record the values of the similarity and sort them to obtain a similarity sorting table; a preset similarity threshold is set, and the keywords exceeding the similarity threshold are selected as similar indicators according to the similarity sorting table, and the key indicators pointed to by the similar indicators are recorded; The key indicator comparison key indicator table based on the similar index direction sends a calling application to the model to which the key indicator table belongs, and the calling application and the key indicator are taken as calling data.

6. The pre-school education key indicator-based AI call operation rule method according to claim 1, characterized in that: The step of transmitting the called data group as reply data to the preschool education AI model through the dynamic data pipeline is: The reply data is stored in the comparison unit, the reply data is encapsulated by the comparison unit and is given an audit watermark layer, the encapsulated reply data is wrapped by the audit watermark layer to obtain encapsulated reply data; The encapsulated reply data is transmitted through the dynamic data pipeline, and a plurality of detection windows are opened on the dynamic data pipeline; When the encapsulated reply data passes through the detection window, the unique ID scratch of the detection window is engraved on the audit watermark layer, and all detection window ID scratches engraved on the audit watermark layer are identified; If the scratch on the audit watermark layer identified does not belong to the detection window of the dynamic data pipeline, the detection window destroys the encapsulated reply data, generates a detection log including the scratch information and sends a retransmission application to the cloud big data center.

7. An AI calling operation rule system based on preschool education key indicators, the AI calling operation rule system based on preschool education key indicators is used to realize the AI calling operation rule method based on preschool education key indicators in any one of claims 1-6, characterized in that: a cloud server module: build a cloud big data center to collect multi-source data of preschool education; a preschool education AI module: build a preschool education AI model and define a calling rule; a data security module: build a dynamic data pipeline, and the cloud big data center and the preschool education AI model interact through the dynamic data pipeline, including: setting data ports on the cloud big data center and the preschool education AI model, and connecting the data ports of the cloud big data center and the preschool education AI model through the dynamic data pipeline; setting a comparison unit in the data port of the cloud big data center, the comparison unit is used to receive the calling data sent by the preschool education AI model, and to perform semantic analysis on the key indicators to retrieve similar mapping indicators to obtain corresponding data keys; the data key is used to access the index album and call the corresponding data group; the called data group is taken as reply data and transmitted to the preschool education AI model through the dynamic data pipeline, and a detection window is set on the dynamic data pipeline, the detection window is used to confirm the safety of the data group.

Images