Method for providing artificial intelligence training platform that automatically matches outsourcing projects by analyzing user data
The AI education platform addresses the challenge of matching educational content and projects by analyzing user data to provide tailored learning and execution paths, enhancing efficiency and success rates through differentiated difficulty levels.
Patent Information
- Application Number
- KR1020250203251
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2025-12-18
- Publication Date
- 2026-11-11
- Estimated Expiration
- 2045-12-18
AI Technical Summary
Conventional AI education platforms fail to comprehensively match educational content and practical projects based on user data, leading to excessive difficulty or simplicity, which can result in decreased learning efficiency, project failures, and operational risks.
An AI education platform that analyzes user data including age, educational background, work experience, content attendance history, and project performance history to match educational content and projects, using multiple AI models to differentiate content and project difficulty levels based on user competency.
The platform ensures a continuous learning path and practical execution path tailored to users, reducing dropout rates, improving learning efficiency, and enhancing project success rates by providing content and projects of appropriate difficulty levels.
Smart Images

Figure 112025143583814-PAT00001_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to a method for providing an artificial intelligence education platform that automatically matches outsourcing projects by analyzing user data. Background Technology
[0002] With the recent rise in interest in artificial intelligence and the proliferation of AI online education platforms (communities), various technologies are being proposed to recommend personalized educational content by analyzing learners' tendencies and achievements, and there is growing interest in technologies that match community members with various AI-related outsourcing projects.
[0003] However, conventional technology was primarily limited to recommending educational content or managing progress, which had the problem of not being sufficiently linked to practical tasks or externally commissioned projects that learners could actually perform.
[0004] Furthermore, even if an outsourcing project matching platform exists, it has limitations in that it fails to comprehensively reflect complex data such as users' content enrollment history, educational attitude scores, educational evaluation scores, and project execution history. This results in matching users with projects that are excessively difficult compared to their capabilities, or conversely, matching them with simple, repetitive projects that do not reflect their potential for growth.
[0005] Furthermore, even for the same educational content or project, learning continuity and the likelihood of success vary depending on the user's educational background, career, age, etc., but there is no structure to separate and determine the type and difficulty level that reflects this, which could lead to operational risks such as increased dropout, decreased learning efficiency, failure to meet project deadlines, and deterioration in the quality of deliverables.
[0006] Therefore, technology is required that can connect education and practical work into a single flow by precisely matching the types and difficulty levels of educational content based on user data, while simultaneously continuously matching the types and difficulty levels of practical projects. Prior Art 1: Registered Patent Publication No. 10-2820719 (June 13, 2025) Prior Art 2: Published Patent Publication No. 10-2024-0127035 (August 22, 2024) Prior Art 3: Published Patent Publication No. 10-2023-0073845 (May 26, 2023) The problem to be solved
[0007] The problem that the present invention aims to solve is to provide a method for providing an artificial intelligence education platform that automatically matches outsourcing projects by analyzing user data.
[0008] The problems that the present invention aims to solve are not limited to those mentioned above, and other unmentioned problems will be clearly understood by a person skilled in the art from the description below. means of solving the problem
[0009] The method of the present invention for solving the above-mentioned problem comprises the steps of: a user terminal transmitting user data to a main server; the main server matching educational content based on the user data using artificial intelligence and transmitting it to the user terminal; the main server matching a project based on the user data using artificial intelligence and transmitting it to the user terminal; the user terminal selecting and taking the matched educational content; and the user terminal selecting and performing the matched project, wherein the user data is characterized as being data regarding the user's age, educational background, work experience, content attendance history, and project performance history. Effects of the invention
[0010] According to the present invention, the main server analyzes the user's age, educational background, work experience, content intake history, and project execution history using artificial intelligence, and matches educational content and projects step-by-step, thereby continuously providing a learning path and a practical execution path suitable for the user.
[0011] In addition, according to the present invention, by operating a first artificial intelligence that generates types of educational content and a second artificial intelligence that outputs the difficulty level of educational content separately, the difficulty level can be differentiated and provided according to the competency level of each user even for the same type of content, thereby preventing excessive learning burden or reduced learning effectiveness.
[0012] In addition, according to the present invention, by operating a third artificial intelligence that generates project types and a fourth artificial intelligence that outputs project difficulty levels separately, a project of appropriate difficulty reflecting the user's practical performance stability and the quality level of deliverables is provided, thereby improving the project execution success rate and delivery deadline compliance rate.
[0013] Furthermore, according to the present invention, since matching reflecting behavioral and performance indicators such as education attitude scores, education evaluation scores, performance attitude scores, and performance evaluation scores is possible, it is possible to achieve the effect of reducing user dropout and improving the reliability and efficiency of platform operation.
[0014] The effects of the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by a person skilled in the art from the description below. Brief explanation of the drawing
[0015] FIG. 1 is a conceptual diagram showing a system constituting the method of the present invention. FIG. 2 is a flowchart showing the steps in which the method of the present invention is performed. Specific details for implementing the invention
[0016] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the present invention, and the present invention is defined only by the scope of the claims.
[0017] The terms used in this specification are for describing embodiments and are not intended to limit the invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. The terms "comprises" and / or "comprising" used in this specification do not exclude the presence or addition of one or more other components in addition to the components mentioned. Throughout the specification, the same reference numerals refer to the same components, and "and / or" includes each of the mentioned components and all combinations of one or more. Although terms such as "first," "second," etc., are used to describe various components, these components are not limited by these terms. These terms are used merely to distinguish one component from another. Therefore, the first component mentioned below may be the second component within the technical scope of the invention.
[0018] Unless otherwise defined, all terms used herein (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.
[0020] Hereinafter, the method (1000) of the present invention will be described with reference to the drawings. FIG. 1 is a conceptual diagram showing a system constituting the method of the present invention, and FIG. 2 is a flowchart showing the steps in which the method of the present invention is performed.
[0021] The method (1000) of the present invention relates to a method of matching and providing educational content and outsourcing projects commissioned from outside based on artificial intelligence in an artificial intelligence education platform (community) through the interaction between a main server (30) and a user terminal (10).
[0022] The method (1000) of the present invention can be performed within a system including a user terminal (10), a network (20), and a main server (30).
[0023] The user terminal (10) may be a terminal used by a member (user) who has joined the platform and is performing artificial intelligence education sessions and outsourcing projects, and the main server (30) may be a server used by the operator of the artificial intelligence education platform (community).
[0024] The user terminal (10) and the main server (30) may be terminals that provide a web page, an app page (app list), a program, or an application related to the method (1000) of the present invention.
[0025] In this case, the user terminal (10) can be implemented as a computer that can access and connect to a main server (30) at a remote location through a network (20).
[0026] In detail, the user terminal (10) is a wireless communication device and may include all kinds of handheld-based wireless communication devices such as navigation, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminal, smartphone, smartpad, tablet PC, etc., as well as wired communication devices such as a general desktop PC, but is not limited thereto.
[0027] A network (20) refers to a connection structure capable of exchanging information between each node, such as a user terminal (10) and a main server (30). Examples of such networks include, but are not limited to, RF, 3GPP (3rd Generation Partnership Project) networks, LTE (Long Term Evolution) networks, 5GPP (5th Generation Partnership Project) networks, WIMAX (World Interoperability for Microwave Access) networks, the Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), Bluetooth networks, NFC networks, satellite broadcasting networks, analog broadcasting networks, and DMB (Digital Multimedia Broadcasting) networks.
[0028] The main server (30) is a server of an operating platform operated by an administrator, and may be a server that provides web pages, app pages, programs, or applications related to the method (1000) of the platform of the present invention.
[0030] The method (1000) of the present invention may include the steps of: a user terminal (10) transmitting user data to a main server (30) (100); the main server (30) matching educational content based on user data using artificial intelligence and transmitting it to the user terminal (10) (200); the main server (30) matching a project based on user data using artificial intelligence and transmitting it to the user terminal (10) (300); the user terminal (10) selecting the matched educational content and taking the course (400); and the user terminal (10) selecting the matched project and performing it (500).
[0031] The step (100) in which the user terminal (10) transmits user data to the main server (30) may be a step in which the user transmits data regarding their profile, content course history, project performance history, etc. to the platform.
[0032] User data may include data regarding the user's age, education level, work experience, content enrollment history, and project performance history, but is not limited to these.
[0033] The user's age may be information that the user directly inputs through the user terminal (10) and may be entered as a number.
[0034] The user's educational background may be data that is automatically generated by setting a grade mapped to the main server (30) when the user directly inputs it through the user terminal (10), and may be classified into high-education and low-education levels.
[0035] The user's career may be data generated by automatically setting a grade mapped to the main server (30) when the user directly inputs it through the user terminal (10), and may be divided into high-experience and low-experience.
[0036] Content course history may include data on the types of completed content (e.g., Python educational content, machine learning educational content, natural language processing educational content, etc.), as well as educational attitude scores and educational evaluation scores, but is not limited to these.
[0037] In this case, the educational attitude score may increase as the user's viewing cycle is shorter, the duration of repeated playback of the same section is longer, the overall content completion rate is higher, and passive viewing behavior patterns (pausing, speeding up playback, background switching) are less frequent. Meanwhile, the educational evaluation score may be the average score on exams taken by the user while performing the educational content.
[0038] Similarly, project execution history may include data on the types of completed projects (e.g., consumer behavior pattern analysis report projects, recommendation score calculation model development projects, etc.), performance attitude scores, and performance evaluation scores, but is not limited to these.
[0039] In this case, the performance attitude score may increase as the user's project execution period is shorter, the time spent on repetitive tasks is longer, the overall project completion rate is higher, and passive performance behavior patterns (work delays, mid-term abandonment, repeated requests for modifications, and delayed communication responses) are less frequent. Meanwhile, the performance evaluation score may be an average score calculated by combining the deliverable quality score, schedule adherence score, and evaluation scores from the client and platform manager that the user received during or after the project execution process.
[0040] The step (200) in which the main server (30) uses artificial intelligence to match educational content based on user data and transmits it to the user terminal (10) may be a step in which user data is analyzed and processed using artificial intelligence on a community platform to select and curate educational content suitable for the user.
[0041] Artificial intelligence may include a first artificial intelligence and a second artificial intelligence. In this case, the first artificial intelligence can generate types of matching educational content (e.g., Python educational content, machine learning educational content, natural language processing educational content, etc.), and the second artificial intelligence can output the difficulty level of the matching educational content (beginner, intermediate, advanced).
[0042] If the content course history of user data is input into the first AI, the types of matching content can be output.
[0043] Specifically, the first artificial intelligence may include: the 1-1 artificial intelligence trained with training data in which the types of educational content that are mutually correlated are set as feature values and the types of educational content that are matched are set as label values; the 1-2 artificial intelligence trained with training data in which the types of educational content that are mutually correlated are set as label values and the types of educational content that are matched are set as feature values and the educational attitude scores are matched are set as label values; the 1-3 artificial intelligence trained with training data in which the types of educational content that are mutually correlated are set as label values and the types of educational content that are matched are set as feature values and the educational evaluation scores are matched are set as label values; and the 1-4 artificial intelligence trained with training data in which the types of educational content that are mutually correlated are set as label values and the types of educational content that are matched are set as label values and the types of educational content that are matched are set as feature values and the types of educational content that are matched are set as label values.
[0044] In this case, the learning data of the first artificial intelligence can be generated based on the content viewing history accumulated from multiple existing users using the community platform. Specifically, the main server (30) can continuously collect and store the type of completed content, the order of content viewing, the time of content viewing, the education attitude score, and the education evaluation score transmitted from the user terminal (10) during the platform operation process.
[0045] Additionally, the main server (30) can further combine the education attitude score and education evaluation score calculated for each completed course content with feature values to form a content transfer relationship that reflects learning attitude and learning performance, rather than a simple content movement relationship, as learning data. This learning data can be configured to reflect whether a specific combination of content actually contributed to learning persistence and performance improvement.
[0046] The main server (30) can match educational content to the user by inputting the type of completed course content and the education attitude score into the first and second artificial intelligence when the user's age is less than the first threshold value and the user's education level is low.
[0047] This is intended to reflect the characteristic that, for users whose age is below the first threshold and who have low educational attainment, attitude toward learning participation has a greater influence on future learning continuation and outcomes than learning achievement. In such user groups, attitudinal factors such as course frequency, repeated playback, and completion rates predict the likelihood of learning success more accurately than test scores or evaluation results; therefore, it is technically advantageous to utilize AI 1 and 2 to match types of educational content that maintain learning interest and engagement without excessively increasing the learning burden.
[0048] In contrast, the main server (30) can match educational content to the user by inputting the type of completed course content and the education evaluation score into the 1st-3rd artificial intelligence when the user's age is greater than or equal to the 1st standard value and the user's education level is low.
[0049] This is intended to reflect the characteristics of users whose age is above the first threshold and who have low educational attainment; while their learning attitude is relatively stable, significant individual differences exist in their actual comprehension and achievement levels. In such cases, adjusting the topics and levels of subsequent content based on objective achievement indicators, such as test scores, rather than diligence in the learning process is effective in preventing learning failure. Therefore, it is reasonable to apply AI 1-3 to match content types based on educational evaluation scores.
[0050] In contrast, the main server (30) can match educational content to the user by inputting the type of completed course content, the education attitude score, and the education evaluation score into the 1st-4th artificial intelligence when the user's age is less than the 1st threshold value and the user's education level is high.
[0051] This is intended to reflect conditions where complex analysis is possible, as both learning attitude and achievement serve as meaningful information when the user's age is below the first threshold and their educational level is high. Since both learning participation patterns and test performance are valid for predicting the likelihood of subsequent learning success in this user group, learning efficiency can be maximized by performing precise matching of educational content types that simultaneously considers attitude and achievement through AI 1-4.
[0052] Furthermore, the main server (30) can match educational content to the user by inputting the type of completed course content into the 1-1 artificial intelligence when the user's age is greater than or equal to the 1st standard value and the user's educational level is high.
[0053] This is intended to enable stable matching of educational content types even when supplementary information, such as user learning attitude scores or educational evaluation scores, is insufficiently accumulated or when the reliability is low due to high variability of such scores. In particular, for users who are highly educated and above the first threshold age, their basic learning comprehension and self-directedness are high, allowing for the establishment of appropriate subsequent learning paths based solely on the logical connections between content types. Therefore, by applying AI 1-1, the intervention of unnecessary variables can be reduced, and the consistency and stability of recommendation results can be ensured.
[0054] If user data's content course history and project execution history are input into the second AI, the difficulty level of the matching content can be output.
[0055] The main server (30) may use the second artificial intelligence to determine the difficulty level suitable for providing the content to the user for the type of educational content matched by the first artificial intelligence.
[0056] The second AI is configured to receive content viewing history and project execution history from user data as input data and output the difficulty level of the matched educational content.
[0057] The second AI may be an AI trained on training data in which the difficulty level of educational content is set as a label value and the content viewing history and project execution history, which are mutually correlated, are matched as feature values. In this case, the training data may be obtained from user data of other past users and data regarding the difficulty level of educational content directly selected by the user, but is not limited thereto.
[0058] In contrast, the second artificial intelligence can receive as input data a content course history consisting of a type of completed course content, an education attitude score, and an education evaluation score, and a project execution history consisting of a type of completed project, an execution attitude score, and an execution evaluation score, among user data transmitted from the user terminal (10) and stored in the main server (30).
[0059] In this case, content course history is used as data reflecting the user's theoretical learning level and learning continuity, and project execution history can be used as data reflecting the user's practical application ability and problem-solving ability.
[0060] The second AI is configured to normalize input data during the preprocessing stage and then calculate content-based learning competency indicators and project-based practical competency indicators, respectively.
[0061] In this case, the content-based learning competency indicator can be calculated by synthesizing the difficulty distribution of completed content, the average value of educational evaluation scores, and the stability of educational attitude scores, and the project-based practical competency indicator can be calculated by synthesizing the difficulty of completed projects, the average value of performance evaluation scores, and the consistency of performance attitude scores.
[0062] The second AI can be configured to calculate the user's overall learning level index (L) using content-based learning competency indicators and project-based practical competency indicators as inputs, and can be calculated in the form "L = h(Ac, Ap)".
[0063] Here, Ac is a content-based learning competency indicator, Ap is a project-based practical competency indicator, and h can be a non-linear function based on an artificial neural network.
[0064] The main server (30) may extract the content difficulty level Dc, the education evaluation score Ec, and the education attitude score Tc for each completed content from the user's content viewing history, and then generate Ac as a weighted average or cumulative statistic for the most recent N contents. At this time, Ac may be generated as a single value or as a vector containing multiple sub-elements, and in one embodiment, Ac may be defined as a vector containing "average test score, average attitude score, average completion rate, average content difficulty" as Ac = [Ec_mean, Tc_mean, Rc_mean, Dc_mean].
[0065] The main server (30) can extract the project-specific difficulty level Dp, performance evaluation score Ep, and performance attitude score Tp from the user's project performance history, and generate Ap by applying a weighted average to the most recent M projects. In one embodiment, Ap can be defined as a vector including "average result score, average attitude score, average project difficulty, and schedule compliance rate" as in Ap = [Ep_mean, Tp_mean, Dp_mean, OnTime_rate], where the schedule compliance rate can be calculated by accumulating values that record whether the project deadline was met in binary.
[0066] Ac, a content-based learning competency indicator, is calculated by synthesizing the difficulty distribution of completed educational content, the average of educational evaluation scores, and the stability of learning attitude scores; it represents the user's theoretical understanding and learning persistence. On the other hand, Ap, a project-based practical competency indicator, is calculated by synthesizing the difficulty of completed projects, performance evaluation scores, performance attitude scores, and schedule adherence; it represents the user's ability to apply knowledge to practical work and problem-solving skills.
[0067] The second artificial intelligence takes Ac and Ap as inputs and calculates a comprehensive learning level index (L) through an artificial neural network-based non-linear function h, where h can be implemented as a neural network structure including multiple hidden layers and activation functions.
[0068] In one embodiment, h may be configured to receive Ac and Ap as inputs, either individually or as vectors, in an input layer, combine Ac and Ap as weights in a first hidden layer and perform a non-linear transformation by an activation function, and combine the transformed values again in a second hidden layer to output L as a single value or an interval value in an output layer.
[0069] For example, when the ReLU or sigmoid function is used as the activation function, the output characteristics can be formed such that L increases gradually in the region where both Ac and Ap are below a certain threshold, and increases rapidly in the region where both Ac and Ap are above a certain threshold.
[0070] Accordingly, even if the content-based learning competency indicator Ac is high, if the project-based practical competency indicator Ap is low, L may be restricted to remain in the intermediate range; conversely, even if Ap is high but Ac is low, L may be controlled so as not to rise to the advanced range.
[0071] As another example, when h is implemented as a multilayer perceptron structure, parameters can be adjusted so that specific combination patterns of Ac and Ap—such as when Ac is at an intermediate level or Ap is very high, or when both Ac and Ap are above an intermediate level with low variability—are recognized as advanced learnable states during the learning process.
[0072] As a result, unlike the simple threshold comparison method, the second AI can flexibly output different L values for various combinations of learning history and execution history.
[0073] Subsequently, the second artificial intelligence can be configured to classify the difficulty level of the matching educational content into beginner, intermediate, or advanced by comparing the comprehensive learning level index (L) with a set value.
[0074] For example, if the overall learning level index (L) is less than the first setting value, the difficulty level of the matching educational content can be set to beginner, if it is greater than the first setting value but less than the second setting value which is greater than the first setting value, it can be set to intermediate, and if it is greater than the second setting value, it can be set to advanced.
[0075] Furthermore, the second AI can output different difficulty levels for the same type of educational content by reflecting users' varying course history and project execution history, thereby enabling the differential provision of the same educational content based on the user's competency level. Accordingly, beginner-level content may be provided to users whose learning level is insufficient, while intermediate or advanced-level content may be provided to users who have accumulated both theoretical learning and practical experience.
[0076] Meanwhile, the learning data of the second artificial intelligence can be generated based on the content viewing history and project execution history accumulated from multiple existing users using the community platform. The main server (30) can construct the learning data by analyzing whether the user continues learning, the project success rate, and whether subsequent high-difficulty content is completed after performing a specific combination of educational content and project, and by setting the results as the correct answer label for the difficulty level of the matching educational content.
[0077] By having the type and difficulty level of educational content determined separately by such a second artificial intelligence, users can receive educational content of a difficulty level that matches both their theoretical learning level and practical performance ability, and achieve the technical effect of preventing excessive learning burden or reduced learning effectiveness.
[0078] According to the above description, in the method (1000) of the present invention, the main server (30) can match educational content with a type and difficulty level suitable for the user based on user data and transmit it to the user terminal (10).
[0079] The step (300) in which the main server (30) uses artificial intelligence to match projects based on user data and transmits them to the user terminal (10) may be a step in which user data is analyzed and processed using artificial intelligence on a community platform to select and curate projects suitable for the user.
[0080] Artificial intelligence may further include a third artificial intelligence and a fourth artificial intelligence. In this case, the third artificial intelligence can generate matching project types (e.g., consumer behavior pattern analysis report project, recommendation score calculation model development project, etc.), and the fourth artificial intelligence can output the matching project difficulty level (beginner, intermediate, advanced).
[0081] If the project execution history of user data is input into the third AI, the type of matching project can be output.
[0082] Specifically, the third artificial intelligence may include the third-1 artificial intelligence trained with training data in which the types of completed projects with mutual correlation are set as feature values and the types of matching projects are set as label values; the third-2 artificial intelligence trained with training data in which the types of completed projects with mutual correlation and performance attitude scores are set as feature values and the types of matching projects are set as label values; the third-3 artificial intelligence trained with training data in which the types of completed projects with mutual correlation and performance evaluation scores are set as feature values and the types of matching projects are set as label values; and the third-4 artificial intelligence trained with training data in which the types of completed projects with mutual correlation, performance attitude scores, and performance evaluation scores are set as feature values and the types of matching projects are set as label values.
[0083] In this case, the training data of the third artificial intelligence can be generated based on the project execution history accumulated from multiple existing users using the community platform. Specifically, the main server (30) can continuously collect and store the type of completed project, the order of project execution, the time of project execution, the execution attitude score, and the execution evaluation score transmitted from the user terminal (10) during the platform operation process.
[0084] Additionally, the main server (30) may further combine the performance attitude score and performance evaluation score calculated for each completed project with feature values to form a content transfer relationship that reflects performance attitude and performance results, rather than a simple project transfer relationship, as training data. This training data may be configured to reflect whether a specific project combination actually contributed to performance continuity and performance improvement.
[0085] The main server (30) can match a project to the user by inputting the type of completed project and the performance attitude score into the 3-2 artificial intelligence when the user's age is less than the first threshold value and the user's experience is low.
[0086] This is intended to reflect the characteristic that, when a user's age is below the first threshold and their experience is low, the stability of the execution process predicts the likelihood of future project completion more strongly than the execution results. In such user groups, while the quality of deliverables can fluctuate significantly depending on external support or task difficulty, performance attitude signals—such as work delays, dropouts, and delayed communication responses—directly reflect the user's self-management ability and potential for continued execution; therefore, it is reasonable to use AI 3-2 to prioritize matching project types with a high probability of completion.
[0087] In contrast, the main server (30) can match educational content to the user by inputting the type of completed project, the performance attitude score, and the performance evaluation score into the 3rd and 4th artificial intelligence when the user's age is greater than or equal to the 1st threshold value and the user's experience is low.
[0088] This is because, in the user group, it is difficult to sufficiently estimate the actual range of performance capabilities based solely on the types of completed projects, and there is a high likelihood that project type transition patterns will be infrequent due to short performance histories. Furthermore, for low-experience users whose age is above the first threshold, the stability of the performance process (performance attitude score), such as performance speed, communication style, and schedule management, as well as the level of deliverable completeness and schedule adherence (performance evaluation score), vary significantly from person to person. Therefore, if only a single score is reflected, there is a risk of over- or under-recommendation. Consequently, it is technically advantageous to use a third or fourth AI that receives both the performance attitude score and the performance evaluation score as feature values in addition to the types of completed projects to make a complex judgment on the project type, as this helps prevent project failure and improves matching accuracy.
[0089] In contrast, the main server (30) can match a project to the user by inputting the type of completed project and the performance evaluation score into the third-third artificial intelligence when the user's age is less than the first threshold value and the user's experience is high.
[0090] This is because the user group possesses high practical proficiency, and thus their attitude during the execution process tends to remain relatively stable; consequently, the additional reflection of performance attitude scores may not significantly increase the discriminative power of recommendation results. On the other hand, for highly experienced users whose age is below the first threshold, the variance in result indicators is relatively large, such as rapidly executing high-difficulty projects or generating high-level deliverables. Furthermore, since the suitability of subsequent project types can be more directly distinguished by performance evaluation scores such as deliverable quality, schedule adherence, and evaluations by clients and managers, applying a third-party AI trained by combining completed project types and performance evaluation scores is advantageous for achieving sophisticated project type matching based on performance.
[0091] Furthermore, the main server (30) can match a project to the user by inputting the type of completed project into the 3-1 artificial intelligence when the user's age is greater than or equal to the first threshold value and the user's experience is high.
[0092] This is because the user group has accumulated a sufficient history of repeatedly executing various projects, allowing the user's potential fields of expertise and subsequent project paths to be reliably derived solely from the correlations and transition patterns between the types of completed projects. Furthermore, in the group of highly experienced professionals, there is a possibility that performance attitude and evaluation scores may converge beyond a certain level or that noise may be introduced due to evaluator bias; consequently, excessively combining score features could distort recommendation results. Therefore, it is technically advantageous to apply AI 3-1, which is trained based on the types of completed projects, to ensure the consistency and reproducibility of recommendations.
[0093] If user data's content course history and project execution history are input into the 4th AI, the difficulty level of the matching project can be output.
[0094] The main server (30) may use the fourth artificial intelligence to determine the difficulty level suitable for providing the project to the user for the type of project matched by the third artificial intelligence.
[0095] The 4th AI is configured to receive content enrollment history and project execution history from user data as input data, and output the difficulty level (beginner, intermediate, advanced) of the matching project.
[0096] The fourth AI may be an AI trained on training data in which intercorrelated content enrollment history and project execution history are set as feature values and the difficulty of the matching project is set as a label value. In this case, the training data may be obtained from user data of other past users and data regarding the difficulty of the project directly selected by the user, but is not limited thereto.
[0097] In contrast, the fourth artificial intelligence can receive as input data a content course history consisting of a type of completed course content, an education attitude score, and an education evaluation score, and a project execution history consisting of a type of completed project, an execution attitude score, and an execution evaluation score, among user data transmitted from the user terminal (10) and stored in the main server (30).
[0098] In this case, content course history can be utilized as data reflecting the user's theoretical learning level and learning continuity, while project execution history can be utilized as data reflecting the user's practical application ability and project execution stability. The 4th AI can be configured to normalize the input data during the preprocessing stage and then calculate content-based project readiness indicators and project-based execution capability indicators, respectively.
[0099] Content-based project readiness indicators can be calculated by synthesizing the difficulty distribution of completed content, the average value of educational evaluation scores, and the stability of educational attitude scores, while project-based performance competency indicators can be calculated by synthesizing the difficulty distribution of completed projects, the average value of performance evaluation scores, the consistency of performance attitude scores, and the schedule compliance rate. Here, the schedule compliance rate can be calculated by accumulating binary records of whether the project deadline was met.
[0100] The fourth AI can be configured to calculate a user's overall project performance level index (P) using content-based project readiness indicators and project-based performance capability indicators as inputs, and the overall project performance level index (P) can be calculated in the form "P = k(Bc, Bp)". Here, Bc is a content-based project readiness indicator, Bp is a project-based performance capability indicator, and k can be a non-linear function based on an artificial neural network.
[0101] The main server (30) may extract the content difficulty level Dc, the education evaluation score Ec, and the education attitude score Tc for each completed content from the user's content viewing history, and then generate Bc as a weighted average or cumulative statistic for the most recent N contents. At this time, Bc may be generated as a single value or as a vector containing multiple sub-elements, and in one embodiment, it may be defined as a vector including the average test score, average attitude score, average completion rate, and average content difficulty, such as Bc = [Ec_mean, Tc_mean, Rc_mean, Dc_mean].
[0102] Additionally, the main server (30) can extract project-specific difficulty level Dp, performance evaluation score Ep, performance attitude score Tp, and schedule compliance data from the user's project performance history, and generate Bp by applying a weighted average to the most recent M projects. In one embodiment, Bp can be defined as a vector including average result score, average attitude score, average project difficulty, and schedule compliance rate, such as Bp = [Ep_mean, Tp_mean, Dp_mean, OnTime_rate].
[0103] The fact that the above k is a non-linear function based on an artificial neural network means that P can be calculated by reflecting the interaction between the two indicators and the threshold change characteristics through a neural network structure that includes multiple hidden layers and activation functions, rather than combining Bc and Bp as a simple sum or average. In one embodiment, k may be configured to receive Bc and Bp as inputs, either individually or in vector form, combine Bc and Bp with weights in the first hidden layer and perform a non-linear transformation by an activation function, and combine the transformed values again in the second hidden layer to output P as a single value or an interval value in the output layer. For example, if a ReLU or sigmoid function is used as the activation function, the output characteristics may be formed such that P increases gradually in the region where both Bc and Bp are below a certain threshold, and P increases rapidly in the region where both Bc and Bp are above a certain threshold.
[0104] Accordingly, even if the content-based project readiness metric Bc is high, if the project-based execution capability metric Bp is low, P can be restricted to remain in the intermediate range; conversely, even if Bp is high but Bc is low, P can be controlled so that it does not rise to the advanced range. Furthermore, if k is implemented as a multi-layer perceptron structure, parameters can be adjusted during the learning process so that specific combination patterns of Bc and Bp—such as cases where Bc is at an intermediate level but Bp is very high, or where both Bc and Bp are above average with low volatility—are recognized as a state capable of executing high-difficulty projects.
[0105] Subsequently, the fourth artificial intelligence can be configured to classify the difficulty of a matching project into beginner, intermediate, or advanced by comparing the comprehensive project performance level index (P) with a set value. For example, if the comprehensive project performance level index (P) is less than the third set value, the difficulty of the matching project can be set to beginner; if it is greater than or equal to the third set value but less than the fourth set value, it can be set to intermediate; and if it is greater than or equal to the fourth set value, it can be set to advanced.
[0106] Furthermore, since the 4th AI can output different difficulty levels for the same project type by reflecting users' varying content enrollment and project execution histories, it can control the provision of the same project differentially based on the user's competency level. Accordingly, beginner-level projects may be provided to users who have not sufficiently developed theoretical learning skills or lack project execution history, while intermediate or advanced-level projects may be provided to users who have accumulated both theoretical learning and project execution experience.
[0107] Meanwhile, the learning data of the fourth artificial intelligence can be generated based on the content viewing history and project execution history accumulated from multiple existing users using the community platform. The main server (30) can construct the learning data by analyzing whether a user who has been matched with a specific project difficulty has successfully completed the project, dropped out midway, or moved on to a higher difficulty project of the same type, and setting the result as the correct label for the matched project difficulty.
[0108] By having the type and difficulty of a project determined separately by this fourth AI, users can receive projects of a difficulty level that matches both their theoretical learning level and project execution ability, thereby preventing excessive project burden or execution failure and achieving the technical effect of improving the project success rate.
[0109] According to the above description, in the method (1000) of the present invention, the main server (30) can match a project with a type and difficulty suitable for the user based on user data and transmit it to the user terminal (10).
[0110] The step (400) in which the user terminal (10) selects and takes the matched educational content and the step (500) in which the user terminal (10) selects and performs the matched project may be a step in which the user checks the matched educational content and project and selects whether to take or perform the educational content or perform the project.
[0112] Although embodiments of the present invention have been described above with reference to the attached drawings, those skilled in the art will understand that the present invention may be implemented in other specific forms without altering its technical concept or essential features. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive.
Claims
Claim 1 A step in which a user terminal transmits user data to a main server; a step in which the main server uses artificial intelligence to match educational content based on the user data and transmits it to the user terminal; a step in which the main server uses artificial intelligence to match a project based on the user data and transmits it to the user terminal; a step in which the user terminal selects the matched educational content and takes the course; The method includes the step of the user terminal selecting and executing a matched project, wherein the user data is data regarding the user's age, educational background, career, content enrollment history, and project execution history, the content enrollment history is data regarding the types of completed content, educational attitude scores, and educational evaluation scores, and the project execution history is data regarding the types of completed projects, execution attitude scores, and execution evaluation scores, and the artificial intelligence includes a third artificial intelligence and a fourth artificial intelligence, and the main server inputs the project execution history of the user data into the third artificial intelligence to output a matched project type, and the third artificial intelligence is a 3-1 artificial intelligence trained with training data in which the types of completed projects with mutual correlation are set as feature values and the types of matched projects are set as label values, a 3-2 artificial intelligence trained with training data in which the types of completed projects with mutual correlation and execution attitude scores are set as feature values and the types of matched projects are set as label values, and a training data trained with the types of completed projects with mutual correlation and execution evaluation scores are set as feature values and the types of matched projects are set as label values. It includes the 3-3 AI and the 3-4 AI trained with training data in which the type of completed project, performance attitude score, and performance evaluation score, which are mutually correlated, are set as feature values, and the type of matching project is set as a label value; and the main server matches a project to the user with the type of matching project output by inputting the type of completed project and the performance attitude score into the 3-2 AI when the user's age is less than the 1st threshold value and the user's experience is low experience.If the user's age is greater than or equal to the first threshold value and the user's experience is low, the type of completed project, performance attitude score, and performance evaluation score are input into the 3rd-4th AI to match the user with the output type of matching project; if the user's age is less than the first threshold value and the user's experience is high, the type of completed project and performance evaluation score are input into the 3rd-3rd AI to match the user with the output type of matching project; if the user's age is greater than or equal to the first threshold value and the user's experience is high, the type of completed project is input into the 3rd-1 AI to match the user with the output type of project; the main server inputs the user data's content enrollment history and project execution history into the 4th AI to output the matching project difficulty; the main server utilizes the 4th AI to determine the difficulty suitable for providing the corresponding project to the user regarding the project type matched by the 3rd AI; and the 4th AI uses the mutually correlated content enrollment history and project execution history as feature values to determine the difficulty of the matching project A method characterized by being an artificial intelligence trained with training data set as a label value.
Citation Information
Patent Citations
A Methods, devices, and systems for selecting and optimizing learning materials based on student feedback in NIE (Newspaper In Education)
KR102657155B1
Method for providing customized education module
KR1020230073845A
Method for supporting project sharing web platform services
KR1020240127035A
System for providing learner-customized educational content
KR102820719B1
