Apparatus and method of user-customized financial education

Abstract

A user-customized financial education device is disclosed, comprising: a memory storing at least one process for providing user-customized financial education; and a processor operating according to said at least one process; wherein the processor provides a first content capable of interacting with the user to collect first response data of the user, analyzes the collected first response data to infer a persona representing the financial propensity of the user, generates a second content based on the inferred persona using a language model, filters the generated second content using a rule-based verification engine based on the user's first response data, and provides the filtered second content as content for user-customized financial education.

Description

Technology Field

[0001] The present invention relates to an artificial intelligence-based financial education provision technology, and more specifically, to an apparatus and method for inferring a user's financial propensity and providing customized education by utilizing a language model. Background Technology

[0002] With the recent rapid increase in retail investors' participation in the financial market, the need for systematic and effective financial education has become a major issue. Particularly amidst the continuous launch of complex and diverse financial products, there is an explosive increase in demand for customized educational services tailored to individuals' levels of financial knowledge and investment propensity.

[0003] In response to these needs, personalized recommendation systems have traditionally been developed, featuring user-customized content delivery capabilities and the collection and analysis of user response data. Existing educational platforms have adopted approaches based on artificial intelligence and machine learning to analyze user profiles, track real-time learning progress and performance, and provide adaptive learning paths. Furthermore, technologies have been utilized to automatically adjust content difficulty by analyzing user behavior patterns and to improve services based on user feedback.

[0004] However, conventional financial education systems have the following fundamental limitations.

[0005] First, existing systems have a critical problem in that they rely on static educational content and fail to reflect real-time changes in financial market conditions. For example, even when market volatility increases rapidly, they continue to provide the same long-term investment strategy education as usual, which can lead to a practical problem where learners cannot develop the ability to respond appropriately to realistic risk situations.

[0006] Second, conventional educational methods are limited to theoretical explanations based on simple text or two-dimensional graphs, making it difficult to intuitively understand complex financial concepts or the practical significance of portfolio construction. In particular, there is a lack of methods for learners to physically experience abstract concepts, such as the correlation between risk and return, which may lead to a discrepancy between theory and practice in actual investment situations.

[0007] Third, the most critical issue is the lack of objective methods to verify the effectiveness of education. Existing systems measure educational outcomes solely through formal indicators, such as quiz accuracy rates or learning completion rates; consequently, there is a fundamental limitation in that there is no way to confirm whether learners' investment decision-making abilities have actually improved and whether this leads to tangible investment results.

[0008] Due to these limitations, there is an urgent technical need for a new type of intelligent financial education system that reflects real-time market conditions, provides an immersive learning environment, and further verifies educational effectiveness by linking it to actual investment performance. Prior art literature

[0009] Korean Patent Publication 10-2023-0174873 (December 29, 2023) The problem to be solved

[0010] The embodiments disclosed in this disclosure disclose a financial education device and method that accurately identify individual financial propensity by analyzing detailed response data of a user and provide customized financial education content generated based thereon.

[0011] The problems that this disclosure 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

[0012] An apparatus according to the present disclosure for achieving the technical problem described above comprises: a memory storing at least one process for providing user-customized financial education; and a processor operating according to the at least one process. The processor provides first content capable of interacting with a user to collect first response data of the user, analyzes the collected first response data to infer a persona representing the financial propensity of the user, generates second content based on the inferred persona using a language model, filters the generated second content using a rule-based verification engine based on the user's first response data, and provides the filtered second content as content for user-customized financial education.

[0013] Meanwhile, the first response data may include exposure-type tendency data including the user's response to the first content and non-exposure-type tendency data including at least one of the user's response speed, whether to stop, and whether to drop off viewing regarding the first content.

[0014] Additionally, the processor may obtain the non-exposure propensity data by sampling a behavioral pattern including at least one of the user's mouse movement, touch pattern, scroll speed, and response time variability regarding the first content in units of 100ms.

[0015] In addition, the rule-based verification engine can extract negative reaction keywords from the exposure-type propensity data and store them as rules, and filter the second content containing the negative reaction keywords based on the stored rules.

[0016] In addition, the processor can infer the persona based on the first response data using a machine learning-based or rule-based multidimensional financial persona inference engine.

[0017] In addition, the above language model can be trained to generate content that reflects financial concepts and values ​​of multiple cultures by including a cultural context adaptation layer.

[0018] Additionally, the processor can collect second reaction data of the user regarding the provided second content and, based on the collected second reaction data, adjust at least one of the difficulty level and topic of the second content in real time.

[0019] In addition, the processor may recommend at least one financial product related to the second content provided.

[0020] In addition, the processor may calculate an optimal difficulty level representing the user's knowledge level using the user's persona and the second response data, and reconstruct and provide the description of the recommended financial product according to the calculated optimal difficulty level.

[0021] Meanwhile, the method according to the present disclosure is a method for providing user-customized financial education performed by a device, comprising: a step of collecting first response data of a user by providing first content capable of interacting with the user; a step of inferring a persona representing the financial propensity of the user by analyzing the collected first response data; a step of generating second content based on the inferred persona using a language model; a step of filtering the generated second content using a rule-based verification engine based on the user's first response data; and a step of providing the filtered second content as content for user-customized financial education.

[0022] In addition to this, a computer program stored on a computer-readable recording medium for executing the present disclosure may be further provided.

[0023] In addition, a computer-readable recording medium for recording a computer program for executing a method for implementing the present disclosure may be further provided. Effects of the invention

[0024] According to the present invention, by capturing and analyzing even subtle user reactions through detailed behavioral pattern sampling, a significant technical effect can be provided to more accurately identify a user's actual financial propensity compared to conventional methods that rely solely on simple surveys or explicit responses. In particular, due to a multidimensional analysis mechanism that combines exposed propensity data and non-exposed propensity data, it is possible to resolve technical challenges regarding a user's potential financial propensity and learning patterns that could not be detected by existing technologies.

[0025] In addition, the content filtering system through the rule-based verification engine of the present invention can provide a significant technical advantage in that it can realize precise content quality management that reflects individual preferences and sensitivities by learning negative response keywords of users and performing personalized filtering.

[0026] In addition, the real-time adaptive difficulty adjustment and topic change function of the present invention analyzes the user's learning progress and responsiveness in real time to dynamically optimize educational content, thereby achieving excellent educational effects that maximize individual learning efficiency and improve learning continuity compared to existing systems that provide static educational courses.

[0027] The effects of the present disclosure 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

[0028] FIG. 1 is a conceptual diagram of a system according to one embodiment of the present disclosure. Figure 2 is a control block diagram of the device shown in Figure 1. FIG. 3 is a flowchart of a financial education method according to one embodiment of the present disclosure. Figure 4 is a detailed flowchart of step S100 shown in Figure 3. Figure 5 is a detailed flowchart of step S300 shown in Figure 3. FIG. 6 is a flowchart of a financial education method according to another embodiment of the present disclosure. Specific details for implementing the invention

[0029] Throughout this disclosure, the same reference numerals denote the same components. This disclosure does not describe all elements of the embodiments, and general content in the art to which this disclosure pertains or content that overlaps between embodiments is omitted. The terms 'part, module, component, block' as used in the specification may be implemented in software or hardware, and depending on the embodiments, a plurality of 'parts, modules, components, blocks' may be implemented as a single component, or a single 'part, module, component, block' may include a plurality of components.

[0030] Throughout the specification, when a part is described as being "connected" to another part, this includes not only cases where they are directly connected but also cases where they are indirectly connected, and indirect connections include connections made via a wireless communication network.

[0031] Furthermore, when it is stated that a part "includes" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0032] Throughout the specification, when it is stated that a component is located "on" another component, this includes not only cases where a component is in contact with another component, but also cases where another component exists between the two components.

[0033] The terms first, second, etc. are used to distinguish one component from another, and the components are not limited by the aforementioned terms.

[0034] Singular expressions include plural expressions unless there is an obvious exception in the context.

[0035] In each step, identification codes are used for convenience of explanation and do not describe the order of the steps; the steps may be performed differently from the specified order unless a specific order is clearly indicated in the context.

[0036] The operating principles and embodiments of the present disclosure will be described below with reference to the attached drawings.

[0037] In this specification, the term "device according to the present disclosure" includes all various devices capable of performing computational processing and providing results to a user. For example, the device according to the present disclosure may include all of a computer, a server device, and a portable terminal, or may be in the form of any one of these.

[0038] Here, the computer may include, for example, a notebook, desktop, laptop, tablet PC, slate PC, etc. equipped with a web browser.

[0039] The above server device is a server that processes information by communicating with an external device, and may include an application server, a computing server, a database server, a file server, a game server, a mail server, a proxy server, and a web server.

[0040] The above portable terminal may include, for example, all types of handheld-based wireless communication devices such as 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) terminals, smartphones, etc., as well as wearable devices such as watches, rings, bracelets, anklets, necklaces, glasses, contact lenses, or head-mounted devices (HMDs).

[0041] Functions related to artificial intelligence according to the present disclosure are operated through a processor and memory. The processor may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. The one or more processors control the processing of input data according to predefined operation rules or artificial intelligence models stored in memory. Alternatively, if the one or more processors are artificial intelligence-dedicated processors, the artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

[0042] The present invention relates to a system that provides personalized financial education content by precisely analyzing a user's financial propensity. Unlike conventional uniform financial education methods, the present invention analyzes multidimensional response data collected through interaction with a user to infer a persona representing an individual's financial propensity, and based on this, can generate personalized educational content by utilizing a language model. Furthermore, the generated content can be filtered through a rule-based verification engine to provide financial education optimized for the user.

[0043] FIG. 1 is a conceptual diagram of a user-customized financial education system according to one embodiment of the present disclosure.

[0044] Referring to FIG. 1, a user-customized financial education system (1) according to the present invention may be configured to include a device (100) and a user terminal (200).

[0045] The device (100) may be a server or cloud-based system that performs the core processing functions of the present invention, and the user terminal (200) may be a computing device of various forms, such as a smartphone, tablet, desktop computer, or laptop. The device (100) and the user terminal (200) may be connected via the Internet, a wireless communication network, or other communication networks to exchange data in real time.

[0046] The overall operating principle of the present invention can be understood as follows. First, the device (100) provides a first content capable of interacting with the user to a user terminal (200) and can collect the user's response from various angles. The collected response data is classified and analyzed into exposure-type tendency data and non-exposure-type tendency data, through which a persona representing the user's financial tendency can be inferred. Based on the inferred persona, a language model generates a customized second content, which is filtered through a rule-based verification engine and can finally be provided to the user.

[0047] FIG. 2 is a conceptual diagram of a user-customized financial education device according to one embodiment of the present disclosure.

[0048] Referring to FIG. 2, the device (100) according to the present invention may be configured to include a processor (110), a communication module (120), a memory (130), a display (140), and an input module (150).

[0049] The processor (110) is a central processing unit responsible for the core processing functions of the present invention and may be composed of, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a NPU (Neural Processing Unit), or a combination thereof, but is not limited thereto. In one embodiment of the present invention, the processor (110) may adopt a multi-core architecture to improve parallel processing performance and may be appropriately selected from, for example, a range of 8 cores to 64 cores.

[0050] The processor (110) may include a plurality of function modules for providing user-customized financial education. These function modules may be implemented in software, in hardware, or in combination of software and hardware.

[0051] The communication module (120) is a component responsible for data communication between the device (100) and the user terminal (200) and can support various communication protocols. In one embodiment of the present invention, the communication module (120) can support protocols such as HTTP / HTTPS, WebSocket, TCP / IP, and UDP, and can preferentially use the WebSocket protocol for real-time bidirectional communication.

[0052] The communication module (120) may also include security functions, such as supporting SSL / TLS encryption, OAuth 2.0 authentication, and JWT (JSON Web Token)-based session management. When transmitting data, the AES-256 encryption algorithm may be used to protect the user's sensitive financial information.

[0053] The memory (130) is a component that stores various data and programs necessary for the operation of the present invention and may include volatile memory and non-volatile memory. Volatile memory may be, for example, DRAM, SRAM, etc., and non-volatile memory may be, for example, flash memory, SSD, HDD, etc.

[0054] The display (140) may be a component that displays a manager interface or a system monitoring screen, and may include, for example, a display panel such as an LCD, OLED, or LED.

[0055] The input module (150) may include input devices such as a keyboard, mouse, and touchpad, and may enable a system administrator to control and monitor the device (100).

[0056] FIG. 3 is a flowchart of a user-customized financial education method according to one embodiment of the present disclosure.

[0057] A user-customized financial education method according to one embodiment of the present disclosure can be performed by the device (100) shown in FIG. 1.

[0058] Referring to FIG. 3, a user-customized financial education method according to one embodiment of the present disclosure may include the step of a processor (110) providing first content (S100), the step of inferring a user's persona (S200), and the step of providing second content based on the persona (S300).

[0059] According to one embodiment of the present invention, the first content is configured to be interactive with a user and may be configured to include at least one type among advertisements, missions, quizzes, and multiple-choice questions as an initial interaction tool for identifying the user's financial propensity.

[0060] Ad types allow for exposing users to advertisements for various financial products or services and observing their reactions. For instance, they can present an advertising spectrum ranging from conservative investment products (e.g., time deposits, government bonds) to aggressive ones (e.g., stocks, cryptocurrencies). Investment propensity can be identified through responses, such as whether users stay on an ad longer, click on it, or move on immediately.

[0061] Mission types can provide financial-related missions that users can directly participate in. For example, users' financial goals and strategies can be identified through missions such as "simulating investment for one week with a virtual 1 million won," "setting an appropriate savings ratio of monthly income," and "planning for retirement funds."

[0062] The quiz type can provide quizzes of varying difficulty levels capable of measuring financial knowledge levels. The quizzes can cover a wide range from basic financial concepts to advanced investment strategies, and can analyze not only the user's accuracy rate but also their behavioral patterns during the problem-solving process.

[0063] Multiple-choice question types can provide various options to identify users' values, preferences, and goals. For example, psychological characteristics can be identified through questions such as, "What do you consider most important when investing?" or "How do you respond when unexpected losses occur?"

[0064] In one embodiment of the present invention, the processor (110) may select an optimal first content based on the user's personal information. The personal information may include at least one of the user's financial habits, consumption patterns, interests, and demographic information.

[0065] For example, the processor (110) can select a first content having a suitable topic by analyzing the user's financial habits, such as existing financial transaction history, savings patterns, and investment experience; consumption patterns, such as the user's card usage history, online shopping patterns, and subscription service usage status; interests, such as the user's online activities, search history, and social media activities; and demographic information, such as age, gender, occupation, income level, and residential area. For example, content related to long-term investment can be provided preferentially to a user with regular saving habits, and content related to household management can be provided preferentially to a user who exhibits impulsive consumption patterns.

[0066] The processor (110) can provide the first selected content to the user through the user terminal (200) (S100).

[0067] The processor (110) can collect user reaction data regarding the first content and infer the user's persona (S200). This will be explained with reference to FIG. 4.

[0068] Figure 4 is a detailed flowchart of step S200 shown in Figure 3.

[0069] Referring to FIG. 4, the step of inferring a user's persona (S200) may consist of collecting exposure-type tendency data (S210), collecting non-exposure-type tendency data (S220), and inferring a persona reflecting these data (S230).

[0070] Exposure-type propensity data may refer to explicit response data consciously provided by the user. For example, exposure-type propensity data may include the user's direct response to the first content, free response, evaluation, and feedback content.

[0071] The processor (110) can collect exposure tendency data for the first content (S210).

[0072] For example, the processor (110) can collect direct response content stored in a structured JSON format, or collect free response content by utilizing a pre-trained language model such as BERT or GPT.

[0073] Non-exposed tendency data may refer to behavioral pattern data that occurs naturally without the user being conscious of it. For example, non-exposed tendency data may include the user's response speed to the first content, whether the viewing was interrupted, and whether the viewing was abandoned.

[0074] The processor (110) can sample the user's behavioral pattern while the first content is being provided and obtain non-exposure tendency data (S220).

[0075] The processor (110) can collect behavioral patterns, including mouse movements, touch patterns, scroll speeds, and response time variability patterns, that occur on the user terminal (200) while the first content is being provided, by sampling them at high resolution in units of 100ms. These behavioral patterns will indicate the user's level of caution, interest, or tendencies.

[0076] For example, the processor (110) can track the user's mouse cursor movement in real time and collect mouse movement patterns including cursor movement speed, linearity of the cursor trajectory, hovering time, and hesitation time before clicking.

[0077] Alternatively, the processor (110) can collect touch patterns including touch pressure, touch duration, multi-touch pattern, touch accuracy, etc.

[0078] Alternatively, the processor (110) can collect scroll speed and patterns including scroll speed, scroll direction change, scroll stop point, scroll acceleration, etc.

[0079] Alternatively, the processor (110) may collect response time variability patterns including the average response time for the interaction question requested in the first content, the standard deviation of the response time, the difference in response time by problem type, and the change in response time by time period.

[0080] Meanwhile, the processor (110) can collect the user's behavioral patterns regarding the first content provided through an API or browser running on the user terminal (200) according to a series of processes including real-time data streaming, data compression and optimization, buffering and batch processing.

[0081] The processor (110) can infer the user's persona by comprehensively reflecting the exposure tendency data and non-exposure tendency data for the first content collected in this way (S230).

[0082] In the following description, the exposure-type propensity data and non-exposure-type propensity data collected for the first content will be referred to as the first response data.

[0083] The processor (110) can infer a persona representing the user's financial propensity by inputting first response data into an artificial intelligence-based persona inference module.

[0084] The AI-based persona inference module can be implemented as a machine learning-based or rule-based multidimensional financial persona inference engine.

[0085] In one embodiment of the present invention, the multidimensional financial persona inference engine may use a Bayesian network-based inference algorithm. A Bayesian network is a probabilistic graph model and may consist of a directed acyclic graph (DAG) representing the conditional independence between variables and a conditional probability distribution of each variable.

[0086] For example, a multidimensional financial persona inference engine can construct a Bayesian network with variables such as Risk Preference (Conservative, Neutral, Aggressive), Investment Horizon (Short-term, Medium-term, Long-term), Financial Knowledge (Beginner, Intermediate, Advanced), Decision Style (Prudent, Intuitive, Analytical), and Loss Aversion (Low, Moderate, High) as nodes. Additionally, a conditional probability table based on the states of the parent nodes can be constructed for each node. This probability table can be trained using expert knowledge, historical data analysis, or machine learning techniques. Furthermore, an inference algorithm can be implemented to calculate the posterior probability of each variable based on the first response data. For instance, Variable Elimination, Junction Tree, or Sampling-based algorithms can be used.

[0087] A multidimensional financial persona inference engine can provide reliability scores along with inference results. For example, reliability can be measured by calculating the entropy of the posterior probability distribution of each variable. Lower entropy (more concentrated the probability distribution) indicates higher reliability.

[0088] In another embodiment of the present invention, the multidimensional financial persona inference engine can build a machine learning-based complementary inference system together with a Bayesian network. For example, it can predict a user's financial propensity from various perspectives using ensemble techniques such as Random Forest, Gradient Boosting, and XGBoost, or discover hidden patterns through recurrent neural networks such as LSTM and GRU by processing user behavior pattern data as a time series, or identify user groups with similar propensity using clustering algorithms such as K-means and DBSCAN, and assign new users to appropriate clusters.

[0089] The processor (110) can provide second content for financial education based on the user persona inferred in this way (S300). This will be explained with reference to FIG. 5.

[0090] Figure 5 is a detailed flowchart of step S300 shown in Figure 3.

[0091] Referring to FIG. 5, the step of providing persona-based second content (S300) may consist of generating second content using a language model (S310), filtering second content using a rule-based validation engine (S320), and providing the filtered second content as financial education content (S330).

[0092] According to one embodiment of the present invention, the second content is a tool configured to provide financial education to a user and may be configured to include at least one type of explanation, simulation quiz, and scenario.

[0093] Explanation types can provide systematic explanations of financial concepts, investment strategies, risk management, and the like. The difficulty level, examples, and presentation style of these explanations can be adjusted to suit the user persona. For instance, explanations emphasizing stability can be provided to users with a conservative inclination, while those emphasizing profitability can be provided to users with an aggressive inclination.

[0094] Simulation quiz types can provide quizzes that allow users to make decisions directly in a simulated environment that mimics actual investment situations. For example, a virtual portfolio may be provided, and users may be asked to select a response strategy based on changes in market conditions, or to compare and analyze the advantages and disadvantages of specific financial products.

[0095] Scenario-based learning presents various financial situations that may actually occur as scenarios, enabling users to learn how to act in those situations. For example, it can provide scenarios such as "response strategies for sudden medical expenses" or "portfolio management during a stock market crash."

[0096] The processor (110) can generate second content by inputting the user's persona into the language model (S310).

[0097] The language model used in the present invention, unlike a general-purpose language model, is a proprietary language model specialized for financial education and may be a lightweight language model (sLLM).

[0098] In an exemplary embodiment of the present invention, the processor (110) may build a large dataset specialized for financial education for learning a language model. For example, the dataset may include financial education data such as existing financial education textbooks, online lecture materials, and educational materials from financial institutions; financial news and analysis data such as financial market news, investment analysis reports, and articles from economic journals; regulatory and legal document data such as financial laws, regulatory guidelines, and compliance guides; and user Q&A data such as financial consultation records, FAQs, and Q&A from online communities.

[0099] The language model used in the present invention includes a cultural context adaptation layer and can generate second content that reflects financial concepts and values ​​of multiple cultures. This may be an important function for providing global services.

[0100] In an exemplary embodiment of the present invention, the processor (110) can database financial terms, concepts, customs, etc. used in each cultural sphere and map cultural financial concepts so that the model can refer to them. For example, it can reflect the difference between the individualistic investment culture of the West and the collectivist savings culture of the East.

[0101] The language model used in the present invention may include a special tokenizer optimized for financial terms and concepts. Unlike a general tokenizer, the financial-specialized tokenizer can process specialized financial terms such as "ROI," "P / E ratio," and "EBITDA" into a single token, appropriately tokenize and process numerical information such as amounts, ratios, and dates, or support currency units and numerical notations of various countries to perform tokenization suitable for global services.

[0102] The language model used in the present invention can be lightweight for on-device execution. This may be important for protecting user privacy and real-time responsiveness.

[0103] The processor (110) can filter the second content generated using such a language model using a rule-based verification engine (S320).

[0104] The processor (110) may include a rule-based verification engine based on the first reaction data.

[0105] The rule-based verification engine used in the present invention can extract negative reaction keywords from exposure-type propensity data included in the first response data and store them as rules. For example, negative reaction keywords may be automatically extracted from keywords expressing negative emotions in a user's text response, extracted through a user's expression of aversion to a specific financial topic or product, or automatically extracted from expressions that may cause discomfort or aversion to the user by building a personalized taboo word database for each user.

[0106] A rule-based verification engine can verify the generated second content through a multi-layered filtering system. For example, it may consist of a first filter that filters out expressions that may be legally problematic, inappropriate investment solicitations, exaggerated return displays, etc., a second filter that filters out the second content containing such content or modifies it with alternative expressions based on negative reaction keywords specific to each user, and a third filter that verifies whether the generated second content aligns with educational purposes and is suitable for the user's current level of knowledge.

[0107] The rules of a rule-based validation engine can be updated in real time, such as by generating new filtering rules based on user feedback, reviews by financial experts, and regulatory changes.

[0108] Meanwhile, the processor (110) can collect user's second reaction data regarding the second content and analyze the second reaction data to control the language model in a feedback loop so as to adjust at least one of the difficulty and topic of the second content in real time.

[0109] The second response data may include exposure-type propensity data and non-exposure-type propensity data collected for the second content, and may be collected in the same manner as the first response data.

[0110] In one embodiment of the present invention, the processor (110) can measure the user's understanding by collecting and analyzing second response data in real time that occurs during the process of the user interacting with the second content. This may include both exposure-type tendency data (correct answer rate, response content, etc.) and non-exposure-type tendency data (response speed, whether interrupted, whether viewing was abandoned, etc.).

[0111] The processor (110) can calculate the accuracy rate for the second content from the exposure tendency data included in the second response data and dynamically adjust the difficulty accordingly. For example, if the accuracy rate is 80% or higher, the difficulty can be increased, and if it is 60% or lower, the difficulty can be decreased.

[0112] The processor (110) can calculate the responsiveness to the second content from the non-exposure type propensity data included in the second response data and change the topic accordingly. For example, the responsiveness can be calculated by combining the concentration based on the time of exit relative to the total time, the engagement based on the number of interactions relative to the total interaction opportunities, and the persistence based on the ratio of completed content.

[0113] Meanwhile, the processor (110) can generate a personalized learning plan by combining real-time second response data regarding the user's persona and second content:

[0114] In one embodiment of the present invention, the processor (110) can analyze the user's response pattern to identify financial areas of relatively low understanding. For example, if the user shows a pattern of answering investment-related questions correctly but frequently getting insurance-related questions wrong, more time can be allocated to insurance education.

[0115] In one embodiment of the present invention, the processor (110) can adjust the method of providing content by identifying the learning method preferred by the user. For example, content containing many charts or graphs can be provided to a user who responds better to visual materials, and detailed document-type content can be provided to a user who prefers text-based explanations.

[0116] FIG. 6 is a flowchart of a user-customized financial education method according to another embodiment of the present disclosure.

[0117] A user-customized financial education method according to another embodiment of the present disclosure may further include a step of recommending financial products based on the learning results of the second content (S400) in addition to the steps of providing first content (S100) illustrated in FIG. 3, inferring a user persona (S200), and providing second content based on the persona (S300). In the following description, only step S400 is described, and the description of the remaining other steps is replaced by the description above.

[0118] The processor (110) can recommend customized financial products by reflecting the learning results of the second content (S400).

[0119] In one embodiment of the present invention, the processor (110) may recommend at least one financial product related to the second content. For example, the processor (110) may receive information on at least one financial product related to the subject of the second content (e.g., time deposits, government bonds, funds, etc.) from an affiliated financial company and recommend it.

[0120] In one embodiment of the present invention, the processor (110) can collect the accuracy rate for the second content from the second response data for the second content, and calculate the portfolio selection freedom by applying the collected accuracy rate to the following mathematical formula 1.

[0121] [Mathematical Formula 1]

[0122] PortfolioFreedom = min(1, AccumulatedScore / MaxScore × WeightFactor)

[0123] In mathematical formula 1, PortfolioFreedom represents the portfolio degrees of freedom, AccumulatedScore represents the cumulative score obtained by the user during a specific learning process of the second content, Maxscore represents the total score obtainable during the learning process, and WeightFactor represents the weight.

[0124] The processor (110) can calculate the AccumulatedScore obtained by the user during a specific learning process according to the second content. For example, the score can be calculated by combining quiz answers, simulation performance, task completion rate, etc., in the second content related to diversified investment.

[0125] The processor (110) can adjust the WeightFactor individually by considering the user's experience level, investment propensity, risk tolerance, etc. For example, based on the user's persona, a pre-set low weight can be applied to a user with a conservative tendency, and a pre-set high weight can be applied to a user with an aggressive tendency.

[0126] The processor (110) can set the number or type of stocks that the user can select among multiple stocks constituting financial products according to the calculated portfolio selection freedom.

[0127] For example, if the degree of freedom of portfolio selection is calculated to be less than 0.3, the processor (110) can recommend a portfolio consisting of basic stocks with high safety by setting it to the beginner level, and can directly select 30% of the number of stocks that make up the portfolio.

[0128] When the degree of freedom of portfolio selection is calculated to be 0.3 or higher and less than 0.7, the processor (110) can set it to an intermediate level and recommend a portfolio of medium risk, and can directly select a number of items corresponding to 30% to 70% of the multiple items constituting the portfolio.

[0129] When the degree of freedom of portfolio selection is calculated to be greater than 0.7, the processor (110) can set it to an advanced level to recommend a portfolio with a high risk and high return risk, and can directly select a number of items corresponding to 70% or more of the multiple items constituting the portfolio.

[0130] In this way, financial products are proposed by reflecting the user's learning results, and learning participation or achievement can be improved by providing differentiated opportunities to directly construct portfolios.

[0131] Meanwhile, the processor (110) can calculate an optimal difficulty level representing the user's knowledge level using the user's persona and second response data, and reconstruct the description of the recommended financial product accordingly.

[0132] The processor (110) can set a knowledge level by comprehensively reflecting the basic knowledge level measured through the first content, the performance in the learning process according to the second content, and the accuracy or depth of the responses to the second content.

[0133] The processor (110) can classify the difficulty levels into beginner, intermediate, and advanced levels according to the knowledge level, and can reconstruct and provide a description of the recommended financial product according to the classified optimal difficulty level.

[0134] For example, at the beginner level, the explanation can be restructured to focus on basic concepts, utilize everyday analogies, and present only key information concisely; at the intermediate level, it can be restructured to include detailed information on the structure and features of the product, comparisons of advantages and disadvantages, and numerical analysis; and at the advanced level, it can be restructured to explain complex financial theories and models, market analysis, and forecasts using specialized terminology.

[0135] Meanwhile, the disclosed embodiments may be implemented in the form of a recording medium that stores instructions executable by a computer. The instructions may be stored in the form of program code and, when executed by a processor, may generate a program module to perform the operation of the disclosed embodiments. The recording medium may be implemented as a computer-readable recording medium.

[0136] Computer-readable recording media include all types of recording media that store instructions that can be decoded by a computer. Examples include ROM (Read Only Memory), RAM (Random Access Memory), magnetic tape, magnetic disk, flash memory, optical data storage devices, etc.

[0137] As described above, the disclosed embodiments have been explained with reference to the attached drawings. Those skilled in the art will understand that the present disclosure may be practiced in forms different from the disclosed embodiments without changing the technical spirit or essential features of the present disclosure. The disclosed embodiments are illustrative and should not be interpreted restrictively. Explanation of the symbols

[0138] 1: Financial education system 100: Device 200: User terminal

Claims

Claim 1 Memory storing at least one process for providing user-customized financial education; and a processor operating according to at least one of the above processes; wherein the processor provides first content capable of interacting with a user to collect first response data of the user, analyzes the collected first response data to infer a persona representing the financial propensity of the user, generates second content based on the inferred persona using a language model, filters the generated second content using a rule-based verification engine based on the user's first response data, and provides the filtered second content as content for user-customized financial education, wherein the rule-based verification engine is composed of a first filter that filters at least one of expressions that may be legal issues, inappropriate investment solicitations, and exaggerated return displays, a second filter that filters the second content containing such content or modifies it with alternative expressions based on negative response keywords specific to each user, and a third filter that verifies whether the generated second content aligns with educational purposes and whether it is suitable for the user's current level of knowledge, and wherein the processor recommends at least one financial product related to the provided second content, wherein Collecting the accuracy rate for the second content from the user's second response data regarding the provided second content, applying the collected accuracy rate to the [Mathematical Formula 1] below to calculate the portfolio selection freedom, and differentially setting and providing the number or ratio of stocks that the user can directly select among a plurality of stocks constituting the financial products recommended in relation to the second content according to the calculated portfolio selection freedom, [Mathematical Formula 1] PortfolioFreedom = min(1, AccumulatedScore / MaxScore × WeightFactor) (where PortfolioFreedom is the portfolio selection freedom,AccumulatedScore is the cumulative score obtained by the user in a specific learning process according to the second content, Maxscore is the total score obtainable in the learning process, and WeightFactor is a weight applied based on whether the user's persona has a conservative or aggressive tendency) User-customized financial education device. Claim 2 A user-customized financial education device according to claim 1, wherein the first response data comprises exposure-type propensity data including the user's response content to the first content and non-exposure-type propensity data including at least one of the user's response speed, whether to stop, and whether to drop out of viewing to the first content. Claim 3 A user-customized financial education device according to paragraph 2, wherein the processor samples a behavioral pattern including at least one of the user's mouse movement, touch pattern, scroll speed, and response time variability regarding the first content in 100ms increments to obtain the non-exposure propensity data. Claim 4 In paragraph 3, the rule-based verification engine extracts negative reaction keywords from the exposure-type propensity data and stores them as rules, and filters the second content containing the negative reaction keywords based on the stored rules, a user-customized financial education device. Claim 5 In claim 1, the processor infers the persona based on the first response data using a machine learning-based or rule-based multidimensional financial persona inference engine, a user-customized financial education device. Claim 6 In claim 1, the language model is a user-customized financial education device trained to generate content reflecting financial concepts and values ​​of multiple cultures, including a cultural context adaptation layer. Claim 7 A user-customized financial education device according to claim 1, wherein the processor collects second response data of the user regarding the provided second content and adjusts at least one of the difficulty level and topic of the second content in real time based on the collected second response data. Claim 8 delete Claim 9 A user-customized financial education device according to claim 1, wherein the processor calculates an optimal difficulty level representing the user's knowledge level using the user's persona and the second response data, and reconstructs and provides explanation content for the recommended financial product according to the calculated optimal difficulty level. Claim 10 A method for providing user-customized financial education performed by a device comprises: a step of providing first content capable of interacting with a user to collect first response data of said user; a step of analyzing said collected first response data to infer a persona representing said user's financial propensity; a step of generating second content based on said inferred persona using a language model; and a step of filtering said generated second content using a rule-based validation engine based on said user's first response data. The method comprises the step of providing the filtered second content as content for the user-customized financial education; wherein the rule-based verification engine comprises a first filter that filters out at least one of expressions that may be legal issues, inappropriate investment solicitations, and exaggerated return displays; a second filter that filters the second content containing such content or modifies it with alternative expressions based on negative reaction keywords specific to each user; and a third filter that verifies whether the generated second content aligns with educational purposes and is suitable for the user's current level of knowledge; and the method further comprises the step of recommending at least one financial product related to the provided second content; wherein the recommendation step comprises: a step of collecting an accuracy rate for the second content from the user's second reaction data regarding the provided second content; and a step of calculating the portfolio selection freedom by applying the collected accuracy rate to [Mathematical Formula 1] below. and a step of differentially setting and providing the number or ratio of stocks that the user can directly select among a plurality of stocks constituting the financial products recommended in relation to the second content, according to the portfolio selection freedom calculated above;A user-customized financial education method including [Mathematical Formula 1] PortfolioFreedom = min(1, AccumulatedScore / MaxScore × WeightFactor) (wherein PortfolioFreedom is the degree of freedom for portfolio selection, AccumulatedScore is the cumulative score obtained by the user in a specific learning process according to the second content, Maxscore is the total score obtainable in the learning process, and WeightFactor is a weight applied based on whether the user's persona has a conservative or aggressive tendency).;

Images