Artificial intelligence-based financial product recommendation method, system, medium and product

Abstract

The application discloses an artificial intelligence-based financial product recommendation method, system, medium and product, and relates to the technical field of data processing.In the method, physical interaction data and underlying configuration parameters of a target user on an intelligent terminal are acquired, and the device operation proficiency and visual cognitive load of the user are evaluated according to the physical interaction data and the underlying configuration parameters; secondly, the operation complexity of a candidate financial product is matched with the proficiency, and products with excessive complexity are removed to obtain a preliminary screening set; subsequently, information extraction and format reconstruction are performed on the description text of the preliminary screening products based on the proficiency and the visual cognitive load, and an accessible display card is generated; finally, the display cards are arranged according to the visual information density, and an aging-adapted financial product recommendation list is generated and output. The technical scheme provided by the application can improve the accuracy of financial product recommendation.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method, system, medium, and product for recommending financial products based on artificial intelligence. Background Technology

[0002] With the deepening of the digital inclusive finance strategy and the widespread adoption of mobile internet technology, smart terminals have become the primary means for financial services to reach users. Banks, insurance companies, securities firms, and other financial institutions use mobile applications to recommend various financial products such as wealth management, loans, and insurance to users, which has become the mainstream model for financial services. Against this backdrop, the ability to accurately recommend suitable financial products based on user characteristics and present them in a way that is easy for users to understand and operate directly determines the accessibility of financial services and the quality of user experience. To improve the accuracy of financial product recommendations and user conversion rates, adopting intelligent recommendation technology to achieve personalized matching and optimized presentation of financial products has become an important trend in the evolution of fintech.

[0003] Currently, existing financial product recommendation methods typically rely on users' historical transaction data, asset status, risk preferences, and other financial attributes to make recommendation decisions. In this model, the system usually filters candidate products with matching risk levels from a product database based on the user's completed risk assessment questionnaire or historical investment behavior. Subsequently, the system uses collaborative filtering or content recommendation algorithms to rank the candidate products according to predicted user preference scores and presents the recommendation results to the user in a standardized display format.

[0004] However, in real-world, diverse user service scenarios, different user groups exhibit significant individual differences and variability in their operational and information processing capabilities when using smart terminals. Existing technologies employ fixed recommendation mechanisms based on financial attributes and standardized display methods, completely detached from users' actual operational capabilities during device use. This fixed recommendation and display mechanism fails to perceive users' actual capacity to handle complex operational processes and their true cognitive load regarding information density, easily leading to situations where recommended financial products are difficult for users to complete the purchase process or understand the product information. Because fixed recommendation logic cannot adapt to complex scenarios with varying user capabilities, recommended products often suffer from hidden operational threshold mismatches, thus reducing the accuracy of financial product recommendations. Summary of the Invention

[0005] This application provides a method, system, medium, and product for recommending financial products based on artificial intelligence, which can improve the accuracy of financial product recommendations.

[0006] The first aspect of this application provides an artificial intelligence-based financial product recommendation method, comprising: Acquire data on the physical interaction behavior of target users on AI terminals and the underlying configuration parameters of the terminal system; Extract the touch dwell time and swipe trajectory jitter features from the physical interaction behavior data, as well as the display font scaling ratio from the underlying configuration parameters; The device operation proficiency and visual cognitive load of the target user are determined based on the touch dwell time, the swipe trajectory jitter characteristics, and the display font scaling ratio. Extract the operational complexity tags of candidate financial products, match and compare the operational complexity tags with the device operation proficiency, and eliminate candidate financial products whose device operation complexity exceeds the standard to obtain the initial screening product set; Based on the proficiency in operating the digital devices and the visual cognitive load, information is extracted and the layout is reconstructed from the product description texts of each financial product in the initial screening product set to generate corresponding accessible display cards. Calculate the visual information density of each of the accessible display cards, arrange the accessible display cards according to the visual information density, generate a recommended list of age-friendly financial products, and output it.

[0007] By employing the aforementioned technical solution, and acquiring physical interaction data of target users on AI terminals and the underlying configuration parameters of the terminal system, and extracting touch dwell time, swipe trajectory jitter features, and display font scaling ratios, it is possible to assess users' device operation proficiency and visual cognitive load based on their actual operational behavior during real-world device use, thereby achieving dynamic perception of users' operational capabilities. Furthermore, by matching the operational complexity tags of candidate financial products with the determined device operation proficiency, financial products with operational complexity exceeding the user's actual capacity can be eliminated during the recommendation process, avoiding the recommendation of complex products that are difficult for users with weak operational capabilities to complete the purchase process. Simultaneously, based on the determined device operation proficiency and visual cognitive load, information is extracted and the layout of the product description texts of each financial product in the initial screening product set is restructured to generate accessible display cards adapted to the user's ability level. By calculating the visual information density of each accessible display card and arranging them accordingly, it is possible to ensure that the display method of the recommendation results matches the user's actual information processing ability, avoiding difficulties in understanding due to excessive information density. Therefore, this application enables the recommendation logic to adapt to scenarios with differences in user operational capabilities, solving the problem of hidden operational thresholds caused by the mismatch between recommended products and users' actual operational capabilities in existing technologies, thereby improving the accuracy of financial product recommendations.

[0008] Optionally, the touchscreen sensor interface of the AI ​​terminal is invoked to obtain the time difference between when the target user's finger touches the screen and when it leaves the screen during a single touch operation, which is used as the touch dwell time; a continuous set of screen coordinate points collected by the touchscreen sensor interface at a preset sampling frequency when the target user performs a swipe operation is obtained; based on the continuous set of screen coordinate points, the swipe direction angle formed by the line connecting two adjacent screen coordinate points is calculated; the angle difference between adjacent swipe direction angles is calculated, and the statistical variance of the angle difference is used as the swipe trajectory jitter feature.

[0009] Optionally, the ratio of the touch dwell time to a preset baseline dwell time is calculated as a sluggishness coefficient, and the ratio of the sliding trajectory jitter characteristics to a preset baseline jitter variance is calculated as a tremor coefficient; the sluggishness coefficient and the tremor coefficient are weighted to obtain an operation resistance index, and the preset baseline proficiency is divided by the operation resistance index to obtain the digital device operation proficiency; the single-screen character capacity of the current smart terminal screen window is calculated according to the display font scaling ratio; the ratio of the preset number of characters in the standard terms of financial products to the single-screen character capacity is calculated to obtain the expected number of page turns; the expected number of page turns is multiplied by the touch dwell time to obtain the visual cognitive load representing the time cost of information acquisition.

[0010] Optionally, based on the visual cognitive load, an upper limit for the number of characters to be retained in the target text is determined; based on a preset financial product dictionary, string matching is performed on the product description text to extract target transaction data, which includes term, amount, and risk level; based on the upper limit for the number of characters to be retained in the target text, a text template of corresponding length is selected from a preset age-friendly expression template library, and the target transaction data is filled into the text template to generate a summary sentence; based on the user's proficiency in operating digital devices, the touch response area of ​​the voice broadcast trigger control is determined; the voice broadcast trigger control and the summary sentence are combined and formatted to encapsulate and generate the accessibility display card.

[0011] Optionally, the interaction nodes of the candidate financial products in the subscription process are analyzed, and the total number of text input characters, the number of sliding verification nodes, and the number of page transition steps are extracted based on the interaction nodes to form the operation complexity label; according to the target user's touch dwell time and the sliding trajectory jitter characteristics, the expected time for the target user to complete a single character input and the expected number of retries to complete a single sliding verification are calculated respectively; the total number of text input characters is multiplied by the expected time to obtain the input estimated time, and the number of sliding verification nodes is multiplied by the expected number of retries and the single sliding baseline time to obtain the verification estimated time; the input estimated time, the verification estimated time, and the page loading time corresponding to the number of page transition steps are summed to obtain the predicted total interaction time for the target user to complete the subscription process; the security session timeout time of the subscription process of the candidate financial products is obtained; when the predicted total interaction time is greater than the security session timeout time, the operation complexity of the candidate financial product is determined to be excessive and the corresponding candidate financial product is removed, and the remaining candidate financial products are retained to form the initial screening product set.

[0012] Optionally, the screen pixel area occupied by each of the accessible display cards and the total number of characters contained therein are statistically analyzed; the ratio of the total number of characters to the screen pixel area is calculated to obtain the visual information density of each of the accessible display cards; the accessible display cards are arranged in ascending order according to the visual information density; the ascending-ordered accessible display cards are rendered onto the screen of the artificial intelligence terminal in a single-column waterfall layout to generate and display the age-friendly financial product recommendation list.

[0013] Optionally, the system collects the continuous touch coordinates of the target user on the age-friendly financial product recommendation list page; when the continuous touch coordinates fall within the display area of ​​the target accessibility display card for a preset duration, and the trajectory dispersion radius formed by the continuous touch coordinates is less than the maximum offset corresponding to the sliding trajectory jitter feature, it determines that the target user has entered a finger-anchored reading state; the system counts the duration of the finger-anchored reading state and compares the duration with the information acquisition time cost corresponding to the visual cognitive load; when the duration is greater than the information acquisition time cost, it determines that the target user has a text decoding lag for the target accessibility display card; the system folds the summary sentences in the target accessibility display card and replaces them with the visual graphics corresponding to the target transaction data, and simultaneously activates the voice parsing and broadcasting for the target transaction data.

[0014] Secondly, embodiments of this application provide an artificial intelligence-based financial product recommendation system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the artificial intelligence-based financial product recommendation system to perform the method described in the first aspect and any possible implementation thereof.

[0015] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on an AI-based financial product recommendation system, cause the AI-based financial product recommendation system to perform the method described in the first aspect and any possible implementation thereof.

[0016] Fourthly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on an AI-based financial product recommendation system, cause the AI-based financial product recommendation system to perform the method described in the first aspect and any possible implementation thereof.

[0017] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: By employing the aforementioned technical solution, and acquiring physical interaction data of target users on AI terminals and the underlying configuration parameters of the terminal system, and extracting touch dwell time, swipe trajectory jitter features, and display font scaling ratios, it is possible to assess users' device operation proficiency and visual cognitive load based on their actual operational behavior during real-world device use, thereby achieving dynamic perception of users' operational capabilities. Furthermore, by matching the operational complexity tags of candidate financial products with the determined device operation proficiency, financial products with operational complexity exceeding the user's actual capacity can be eliminated during the recommendation process, avoiding the recommendation of complex products that are difficult for users with weak operational capabilities to complete the purchase process. Simultaneously, based on the determined device operation proficiency and visual cognitive load, information is extracted and the layout of the product description texts of each financial product in the initial screening product set is restructured to generate accessible display cards adapted to the user's ability level. By calculating the visual information density of each accessible display card and arranging them accordingly, it is possible to ensure that the display method of the recommendation results matches the user's actual information processing ability, avoiding difficulties in understanding due to excessive information density. Therefore, this application enables the recommendation logic to adapt to scenarios with differences in user operational capabilities, solving the problem of hidden operational thresholds caused by the mismatch between recommended products and users' actual operational capabilities in existing technologies, thereby improving the accuracy of financial product recommendations. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the artificial intelligence-based financial product recommendation method disclosed in the embodiments of this application; Figure 2 This is another flowchart illustrating the artificial intelligence-based financial product recommendation method disclosed in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of a system provided in an embodiment of this application.

[0019] Explanation of reference numerals in the attached drawings: 301, Central Processing Unit; 302, Read-Only Memory; 303, Random Access Memory; 304, Bus; 305, Input / Output Interface; 306, Input Section; 307, Output Section; 308, Storage Section; 309, Communication Section; 310, Driver; 311, Removable Media. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0021] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0022] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0023] This application provides a method for recommending financial products based on artificial intelligence, referring to... Figure 1 , Figure 1This is a flowchart illustrating an artificial intelligence-based financial product recommendation method provided in an embodiment of this application. The method is applied to a system, which refers to a hardware and software integrated platform capable of executing an artificial intelligence-based financial product recommendation program. The system can execute an artificial intelligence-based financial product recommendation program, and the method includes steps 101 to 106, as follows: Step 101: Obtain the target user's physical interaction behavior data on the AI ​​terminal and the underlying configuration parameters of the terminal system.

[0024] In this embodiment, the target user refers to the actual user who uses the AI ​​terminal to browse and operate financial products, particularly a diverse user group that may include special groups such as the elderly. The AI ​​terminal refers to an electronic device with intelligent processing capabilities, such as smartphones, tablets, smart wearable devices, and other terminal devices capable of running financial applications. Physical interaction behavior data refers to various interactive action data generated by the user during operation of the AI ​​terminal, including but not limited to touch screen methods, pressure levels, dwell time, and swipe trajectories—behavioral data that can be captured by the terminal device's sensors. Low-level configuration parameters represent system-level settings of the AI ​​terminal, such as display parameter configurations, system auxiliary function settings, and interface customization parameters—low-level system parameters that affect the user's interactive experience.

[0025] Specifically, by calling the touchscreen sensor interface, accelerometer sensor interface, and system configuration management interface of the AI ​​terminal, real-time physical interaction data and system configuration status of the target user during the operation of the terminal device are collected. First, touchscreen sensor events are monitored, and raw interaction data such as the timestamp, coordinate position, and pressure value of the user's finger contacting the screen are recorded. Second, the system-level configuration information currently set by the user is read from the terminal operating system, including low-level parameters such as display font size, contrast settings, and accessibility function activation status. Then, the collected physical interaction behavior data and low-level configuration parameters are preliminarily processed and formatted, and uniformly converted into a standard data format. Finally, the processed data is saved to a temporary data cache area to prepare for subsequent analysis and feature extraction.

[0026] Step 102: Extract the touch dwell time and swipe trajectory jitter features from the physical interaction behavior data, as well as the display font scaling ratio from the underlying configuration parameters.

[0027] In this embodiment, touch dwell time refers to the duration of a user's finger on the touch screen from contact to removal, representing the duration and reaction speed of the user's touch operation. Slide trajectory jitter indicates the stability of the finger trajectory during a swiping operation, typically quantified by the degree of change in the swiping direction angle, reflecting the user's fine control over their hand movements. Display font scaling refers to the text display size ratio parameter set by the user in the terminal system, usually presented as a percentage of the standard font size, representing the user's personalized needs for text size and their visual recognition ability.

[0028] Specifically, firstly, user touch event records are filtered from the acquired physical interaction behavior data. The start and end timestamps of each touch operation are extracted, and the difference between the two is calculated to obtain the touch dwell time value, forming a touch dwell time sequence data. Then, for user swipe operation events, a continuous coordinate point sequence recorded at a preset sampling frequency during the swipe is extracted. For any two adjacent points, the swipe direction angle is calculated, i.e., the angle between the line connecting the two points and the horizontal direction. Further, the difference sequence between adjacent swipe direction angles is calculated, and the statistical variance of this difference sequence is calculated to obtain the jitter feature value characterizing the stability of the finger swipe trajectory. Simultaneously, the display font scaling ratio value set by the system is read from the underlying configuration parameters, usually expressed as a percentage value relative to the standard font size.

[0029] In one possible implementation, the touch duration and swipe trajectory jitter features are extracted from the physical interaction behavior data, specifically including steps 1021-1023, as follows: Step 1021: Call the touch screen sensor interface of the AI ​​terminal to obtain the time difference between when the target user's finger touches the screen and when it leaves the screen during a single touch operation, and use this as the touch dwell time.

[0030] In this embodiment, touch dwell time refers to the duration of contact between a user's finger and the touchscreen, representing the length of time a user performs a touch operation. An AI terminal refers to an electronic device with intelligent processing capabilities, such as a smartphone or tablet equipped with a touchscreen. A touchscreen sensor interface refers to a system-level interface in the device used to receive and process touch events, capable of capturing various physical parameters of the user's finger contacting the screen. A single touch operation represents the complete touch process from when the user's finger touches the screen to when it leaves the screen, including the two basic actions of pressing and lifting. The time difference is used to represent the interval between two points in time; in this application scenario, it specifically refers to the difference between the touch start time and the touch end time.

[0031] Specifically, the system first calls the touchscreen sensor interface of the AI ​​terminal via a system-level API to register a touch event listener, capturing user interaction events with the touchscreen in real time. When a touch down event (touchDown or touchStart) is detected, the current system timestamp T1 is recorded, accurate to the millisecond level. The system continues listening for touch events, and when a touch up event (touchUp or touchEnd) is detected, the current system timestamp T2 is recorded again. The difference between T2 and T1 is calculated to obtain the duration of the touch operation. For multi-touch scenarios, different touch points are distinguished by their unique identifiers (usually pointerId), and the duration of each touch point is calculated separately. For multiple consecutive touch operations, the duration of each touch is calculated separately, and statistical indicators such as average, maximum, or minimum values ​​can be calculated as needed to quantify the user's touch behavior characteristics.

[0032] Step 1022: Obtain the set of continuous screen coordinate points collected by the touch screen sensor interface at a preset sampling frequency when the target user performs a swipe operation.

[0033] In this embodiment, the continuous screen coordinate point set refers to a series of finger position coordinate data collected by the touchscreen sensor at certain time intervals during the user's screen swiping process. The swiping operation represents the interaction method where the user's finger remains in contact with and moves on the touchscreen; it is a basic action common in smart devices such as browsing, page turning, and dragging. The preset sampling frequency is used to represent the time interval for the system to acquire touchscreen coordinate points, usually in Hertz (Hz), representing the number of samples per second. A screen coordinate point is a two-dimensional point representing a position in the device's screen coordinate system, typically composed of two components: a horizontal coordinate (x) and a vertical coordinate (y), used to accurately locate the user's touch position on the screen.

[0034] Specifically, the system first calls the touchscreen sensor interface's event listening mechanism via the system API to register callback functions for touch movement events (touchMove). An appropriate sampling frequency parameter is set, typically between 60Hz and 120Hz, meaning coordinate data is collected every 8.33 milliseconds to 16.67 milliseconds to balance data accuracy and system resource consumption. When a user initiates a swipe, the swipe event listener is triggered, recording the screen coordinates (x, y) of the touch point at each sampling moment, where x represents the horizontal pixel position and y represents the vertical pixel position. The collected coordinate points are organized into an ordered sequence P = {P1, P2, ..., Pn} in chronological order, where Pi = (xi, yi) represents the coordinates of the i-th sampling point, and n represents the total number of coordinate points collected during the entire swipe. For multi-finger swipe scenarios, the trajectories of different fingers are distinguished by touch point IDs, constructing multiple coordinate point sequences for each. Simultaneously, the timestamp corresponding to each coordinate point is recorded, forming a time-position mapping relationship, providing a data foundation for subsequent calculations of derived features such as swipe speed and acceleration.

[0035] Step 1023: Based on the continuous set of screen coordinate points, calculate the sliding direction angle formed by the line connecting two adjacent screen coordinate points; calculate the angle difference between adjacent sliding direction angles, and use the statistical variance of the angle difference as the feature of sliding trajectory jitter.

[0036] In this embodiment, the swipe trajectory jitter feature refers to a numerical indicator used to quantify the smoothness of a user's swipe operation, representing the stability and linearity of the swipe trajectory. The swipe direction angle represents the angle between the line connecting two adjacent screen coordinate points and the horizontal direction, expressed in radians or angles, and describes the direction of movement of the swipe trajectory in each small segment. The angle difference refers to the magnitude of the difference between two adjacent swipe direction angles, reflecting the drastic change in swipe direction. Statistical variance is a measure of the dispersion of a set of data; in this scenario, it is used to quantify the fluctuation of the angle difference sequence. A larger variance value indicates a less stable swipe trajectory.

[0037] Specifically, firstly, based on the acquired continuous screen coordinate point set P={P1, P2, ..., Pn}, for any two adjacent points Pi(xi, yi) and Pi+1(xi+1, yi+1), the angle between the line connecting them and the positive horizontal direction is calculated, i.e., the sliding direction angle θi. The formula for calculating the sliding direction angle is: θi=arctan((yi+1-yi) / (xi+1-xi)). When xi+1=xi, θi=π / 2 or θi=-π / 2 is taken according to the relationship between yi+1 and yi. The direction angles calculated for all adjacent point pairs are combined into a sequence Θ={θ1, θ2, ..., θn-1}. Then, the difference between adjacent direction angles is calculated to obtain the angle difference sequence Δθ={δ1, δ2, ..., δn-2}, where δi=θi+1-θi. To handle angle jumps between 0 and 2π, each difference δi is normalized to ensure its range is within [-π, π]. Finally, the statistical variance σ² of the angle difference sequence Δθ is calculated using the formula: σ² = (1 / (n-2))·Σ(δi-μ)², where μ is the average value of the angle difference sequence, μ = (1 / (n-2))·Σδi, and i ranges from 1 to n-2. The statistical variance σ² represents the jitter characteristic of the sliding trajectory. A larger value indicates a more drastic change in the direction of the user's sliding trajectory and a higher degree of jitter; a smaller value indicates a sliding trajectory closer to a straight line and a smoother operation.

[0038] Step 103: Determine the target user's device operation proficiency and visual cognitive load based on touch dwell time, swipe trajectory jitter characteristics, and display font scaling ratio.

[0039] In this embodiment, device operation proficiency refers to the user's skill level and operational fluency in using a smart terminal device, representing the user's ability to complete various interactive tasks. Visual cognitive load represents the psychological resource consumption required by the user when processing visual information presented on the screen, reflecting the ease or difficulty and time cost of receiving, processing, and understanding visual information. Device operation proficiency and visual cognitive load together constitute the evaluation dimensions of the user's digital ability and are key parameters for subsequent product matching and display optimization.

[0040] Specifically, firstly, the ratio of touch dwell time to a preset baseline dwell time is calculated to obtain the sluggishness coefficient, which is the average touch dwell time actually measured by the user divided by the system's preset standard user touch dwell time reference value. Then, the ratio of the sliding trajectory jitter characteristic to a preset baseline jitter variance is calculated to obtain the tremor coefficient, which is the measured trajectory jitter variance divided by the system's preset standard tremor variance reference value. Next, the sluggishness coefficient and tremor coefficient are weighted according to preset weights to obtain the operation resistance index, with the total weight coefficients being 1. The preset baseline proficiency is divided by the operation resistance index to obtain the device operation proficiency. For the calculation of visual cognitive load, firstly, the single-screen character capacity of the current smart terminal screen is calculated based on the display font scaling ratio, which is the reference character capacity at the standard font size divided by the square of the scaling ratio. Then, the ratio of the preset number of characters in the standard terms of financial products to the single-screen character capacity is calculated and rounded up to obtain the expected number of page turns. Finally, the expected number of page turns is multiplied by the touch dwell time to obtain the visual cognitive load, which represents the time cost of information acquisition.

[0041] In one possible implementation, the device operation proficiency and visual cognitive load of the target user are determined based on the touch dwell time, swipe trajectory jitter characteristics, and display font scaling ratio. Specifically, this includes steps 1031-1034, as follows: Step 1031: Calculate the ratio of the touch dwell time to the preset baseline dwell time as the sluggishness coefficient, and the ratio of the sliding trajectory jitter characteristics to the preset baseline jitter variance as the vibration coefficient.

[0042] In this embodiment, the sluggishness coefficient is the ratio of the actual user touch operation time to the standard operation time, used to represent the degree of delay in the user's operation speed. Touch dwell time represents the duration of contact between the user's finger and the screen, reflecting the time required for the user to complete a basic touch operation. The baseline dwell time refers to a system-preset reference time value for a standard user to complete a touch operation, used as a comparison benchmark. The jitter coefficient is the ratio of the degree of jitter in the user's sliding trajectory to the standard jitter degree, used to quantify the degree of deviation of the user's finger stability. The sliding trajectory jitter characteristic represents the severity of directional changes during the user's sliding operation, usually expressed as the statistical variance of angle changes. The baseline jitter variance refers to a system-preset reference value for the trajectory jitter of a standard user's sliding operation, used as a comparison benchmark.

[0043] Specifically, firstly, the system obtains the target user's touch dwell time value and simultaneously reads a preset baseline dwell time from the system configuration. This value is typically based on the average operation time of normal users derived from a large amount of statistical data. The ratio of these two values ​​is calculated, known as the slack coefficient. A slack coefficient greater than 1 indicates that the user's touch operation time is longer than the standard time; a slack coefficient less than 1 indicates that the user's touch operation time is shorter than the standard time; and a slack coefficient equal to 1 indicates that the user's touch operation time is the same as the standard time. Next, the system obtains the target user's swipe trajectory jitter characteristic value, i.e., the statistical variance of the angle difference, and simultaneously reads a preset baseline jitter variance from the system configuration. This value is also based on the average jitter level of normal user swipe trajectories derived from statistical data. The ratio of these two values ​​is calculated, known as the tremor coefficient. A tremor coefficient greater than 1 indicates that the user's swipe trajectory jitter level is higher than the standard level; a tremor coefficient less than 1 indicates that the user's swipe trajectory jitter level is lower than the standard level; and a tremor coefficient equal to 1 indicates that the user's swipe trajectory jitter level is the same as the standard level.

[0044] Step 1032: Perform a weighted calculation on the sluggishness coefficient and the flutter coefficient to obtain the operating resistance index, and divide the preset benchmark proficiency by the operating resistance index to obtain the digital equipment operating proficiency.

[0045] In this embodiment, the operational resistance index is a quantitative indicator of the degree of operational obstacle obtained by comprehensively considering the user's touch speed and finger stability. It is used to represent the overall difficulty faced by the user in completing device interaction operations. Weighted calculation refers to a method of comprehensively calculating multiple factors according to different levels of importance, adjusting the influence ratio of each factor by assigning weight coefficients. Digital device operation proficiency refers to the user's skill level and fluency in using smart terminal devices, usually presented as a score or level, reflecting the user's ability to complete various digital interaction tasks. Benchmark proficiency represents a system-preset standard operation level reference value, used to establish a comparison benchmark for proficiency assessment.

[0046] Specifically, firstly, weighting parameters for the sluggishness coefficient and the tremor coefficient are set, with the sum of these two parameters equal to 1. These parameters are used to adjust the relative importance of touch speed and finger stability in the evaluation process. Based on the calculated sluggishness coefficient and tremor coefficient, a weighted average method is used to calculate the operation resistance index, which is the sum of the sluggishness coefficient multiplied by its weight and the tremor coefficient multiplied by its weight. The larger the operation resistance index, the more obstacles the user faces when operating the smart terminal device; the smaller the operation resistance index, the smoother the user's operation. Then, a preset baseline proficiency value is obtained from the system configuration. This value represents the standard user's operational proficiency under ideal conditions. The quotient of the baseline proficiency divided by the operation resistance index is calculated, which is the digital device operation proficiency. When the operation resistance index is greater than 1, the digital device operation proficiency is less than the baseline proficiency, indicating that the user's operation ability is below the standard level; when the operation resistance index is less than 1, the digital device operation proficiency is greater than the baseline proficiency, indicating that the user's operation ability is above the standard level; when the operation resistance index is equal to 1, the digital device operation proficiency is equal to the baseline proficiency, indicating that the user's operation ability is the same as the standard level.

[0047] Step 1033: Calculate the single-screen character capacity of the current smart terminal screen window based on the display font scaling ratio.

[0048] In this embodiment, single-screen character capacity refers to the number of text characters that a smart terminal screen can display at one time, representing the screen's information carrying capacity. The display font scaling ratio represents the proportion of text size adjusted by the user in the system settings, usually presented as a percentage, such as 100% representing the standard size and 150% representing a 1.5x magnification. The smart terminal screen window refers to the visible area of ​​the device's displayed content; its size is determined by the screen's physical dimensions and resolution, and it is the interface range for the user to obtain visual information. The standard character size refers to the screen space occupied by a single character in the system's default state, usually measured in pixels. The reference character capacity represents the number of characters that the screen can display at the standard font size (scaling ratio of 100%), and is the basic value for calculating the actual character capacity.

[0049] Specifically, first, the display font scaling ratio value in the system settings is obtained. This value is usually expressed as a percentage relative to the standard font size, such as 100%, 125%, 150%, etc., which translates to 1.0, 1.25, 1.5, etc. in decimal form. Then, the reference character capacity of the current smart terminal at the standard font size is obtained from the system configuration. This value represents the maximum number of characters the screen can display without font scaling. The calculation of the reference character capacity considers the effective display area of ​​the screen and the average area occupied by a standard character, and is related to the device's screen size and resolution. Since font scaling changes the screen space occupied by a single character, the number of characters the screen can accommodate decreases when the font is enlarged. Character capacity is inversely proportional to font size, and since characters expand in both horizontal and vertical directions, it is inversely proportional to the square of the font scaling ratio. Based on this relationship, the single-screen character capacity at the current font scaling ratio is calculated by dividing the reference character capacity by the square of the display font scaling ratio.

[0050] Step 1034: Calculate the ratio of the preset number of characters in the standard terms of financial products to the character capacity of a single screen to obtain the expected number of page turns; multiply the expected number of page turns by the touch dwell time to obtain the visual cognitive load that represents the time cost of information acquisition.

[0051] In this embodiment, the expected number of page turns refers to the number of screen flips required for a user to read the complete content, representing the number of interactions during content browsing. The character count of standard financial product terms represents the total number of characters in the text describing the characteristics and rules of the financial product, serving as a fundamental indicator of information complexity. Visual cognitive load refers to the psychological resource consumption required for a user to process visual information presented on the screen, typically quantified by time cost, reflecting the ease with which a user receives, processes, and understands visual information. Information acquisition time cost represents the total time required for a user to fully acquire and understand specific information content, and is an important indicator of information accessibility. Page turning refers to the interactive action of switching screen content through touch swiping or other methods, a necessary operation when browsing long content.

[0052] Specifically, first, the preset number of characters in the standard terms and conditions of the financial product is obtained. This value represents the total number of text characters required to fully describe the financial product. Then, based on the calculated single-screen character capacity, the ratio of the preset number of characters in the standard terms and conditions to the single-screen character capacity is calculated. Since the number of page turns must be an integer, and the last page may not fill a full screen, the calculation result needs to be rounded up to obtain the expected number of page turns. For example, if the total number of characters in the standard terms and conditions of the financial product is 5000, and the single-screen character capacity is 600, then the expected number of page turns is 9. Next, the target user's touch dwell time is obtained. This value reflects the time required for the user to complete a single touch operation (such as turning a page). Multiplying the expected number of page turns by the touch dwell time yields the visual cognitive load, which represents the time cost of information acquisition. For example, if the expected number of page turns is 9, and the user's touch dwell time is 2 seconds, then the visual cognitive load is 18 seconds, indicating that the estimated time cost for the user to fully browse the financial product information is 18 seconds.

[0053] Step 104: Extract the operational complexity labels of candidate financial products, match and compare the operational complexity labels with the proficiency of equipment operation, and eliminate candidate financial products whose equipment operation complexity exceeds the standard to obtain the initial screening product set.

[0054] In this embodiment, the operational complexity label refers to a set of multi-dimensional indicators used to quantify the complexity of the financial product subscription process, including dimensions such as text input volume, number of verification operations, and page navigation steps. Candidate financial products represent various financial service products offered by financial institutions for users to choose from, such as deposit products, wealth management products, insurance products, and fund products. The initial screening product set refers to a subset of financial products suitable for the target user's operational ability level after operational complexity matching; it serves as the foundational product pool for subsequent display optimization.

[0055] Specifically, the operational process of each candidate financial product is first analyzed, examining various interaction nodes in the subscription process and extracting three key indicators: total number of text input characters (including account information, amount, verification code, etc., the total number of characters that users need to input), number of sliding verification nodes (the number of nodes that users need to complete, such as slider verification and graphic verification), and number of page transition steps (the number of pages that need to be traversed from product browsing to completing the purchase). These three indicators are combined to form an operational complexity label. Then, based on the calculated target user touch dwell time and sliding trajectory jitter characteristics, the expected time for a user to complete a single character input and the expected number of retries to complete a single sliding verification are calculated, obtained through corresponding mapping functions. Multiplying the total number of text input characters by the expected time gives the estimated input time; multiplying the number of sliding verification nodes by the expected number of retries and the baseline time for a single sliding gives the estimated verification time. Simultaneously, based on the number of page transition steps and the system's preset average page loading time, the total page loading time is calculated. The predicted total interaction time for a target user to complete the subscription process is obtained by summing the estimated input time, the estimated verification time, and the total page loading time. The security session timeout for each candidate financial product's subscription process is obtained. If the predicted total interaction time exceeds the security session timeout, the operational complexity of that candidate financial product is deemed excessive, and it is removed from the candidate product set. The remaining candidate financial products constitute the initial screening product set.

[0056] In one possible implementation, operational complexity labels of candidate financial products are extracted, and these labels are matched and compared with the proficiency of equipment operation. Candidate financial products whose operational complexity exceeds the standard are eliminated to obtain a preliminary set of products. This process includes steps 1041-1045, as follows: Step 1041: Analyze the interaction nodes of the candidate financial products in the subscription process, and extract the total number of text input characters, the number of sliding verification nodes, and the number of page transition steps based on the interaction nodes, and combine them to form an operation complexity label.

[0057] In this embodiment, an interaction node refers to a key point where a user needs to interact with a smart terminal device during the financial product subscription process, representing the various operational steps a user must go through to complete the subscription. Candidate financial products refer to investment and wealth management products recommended to target users after preliminary screening, including but not limited to funds, wealth management products, and insurance. The subscription process refers to the complete operational process a user goes through from selecting a financial product to completing the purchase, typically including multiple steps such as information browsing, identity verification, parameter setting, and order confirmation. The total number of text input characters represents the total number of characters a user needs to manually input throughout the subscription process, including account passwords, amount amounts, and personal information. The number of sliding verification nodes refers to the number of security verification points in the subscription process that require the user to complete through sliding operations, such as slider verification codes and image verification. The number of page transition steps represents the number of page switches required to complete the entire subscription process, reflecting the complexity and depth of the process. The operation complexity label refers to the financial product operation difficulty assessment result composed of the above multiple indicators, used to quantify the complexity of the product interaction process.

[0058] Specifically, the subscription process of the target candidate financial product is first systematically analyzed. By parsing its application interface or webpage structure, all nodes that users need to interact with during the subscription process are identified. Interaction node identification uses an interface element scanning method to detect all controls requiring user input or operation, such as text boxes, buttons, and sliders. Then, based on the identified interaction nodes, three types of key indicator data are extracted: First, the preset character length limit of all text input boxes in the subscription process is calculated, and the total number of text input characters required to complete the subscription is calculated based on the actual requirements of the required fields. Second, all sliding verification components included in the process, such as slider verification codes and puzzle verification, are detected, and the number of sliding verification nodes is calculated. Finally, by analyzing the logical structure of the subscription process, the number of pages traversed from start to finish is counted, resulting in the page transition steps. These three indicator data are organized in a structured manner to form an operation complexity label describing the interaction complexity of the financial product. This label is represented in the form of <total number of text input characters, number of sliding verification nodes, number of page transition steps>, and is used for subsequent operation difficulty assessment calculations.

[0059] Step 1042: Based on the target user's touch dwell time and swipe trajectory jitter characteristics, calculate the expected time for the target user to complete a single character input and the expected number of retries to complete a single swipe verification.

[0060] In this embodiment, expected time refers to the time predicted based on a user's historical operation characteristics for completing a specific interactive task, used to quantify individual differences in user operation efficiency. Touch dwell time represents the duration of contact between the user's finger and the screen, reflecting the time required for a user to complete a basic touch operation. Swipe trajectory jitter refers to the irregularity and deviation of the trajectory when a user performs a swipe operation, typically expressed as the statistical variance of angle changes, reflecting the stability of the user's fingers. Expected time for single character input represents the average time a specific user is expected to take to complete a single character input operation, serving as a fundamental indicator for evaluating text input efficiency. Expected retries for completing a single swipe verification refers to the average number of attempts a user needs to successfully complete a swipe verification, reflecting the difficulty of performing fine touch operations.

[0061] Specifically, firstly, using the acquired target user touch dwell time data and combining it with the system's preset standard character input baseline time, the expected time for a single character input by the user is calculated proportionally. The calculation formula is: Expected time for a single character input = Standard character input baseline time × (Target user touch dwell time ÷ Standard touch dwell time). Here, the standard character input baseline time and standard touch dwell time are both preset baseline parameters representing the average user's operation time. Then, based on the acquired swipe trajectory jitter characteristics, i.e., the statistical variance of the angle difference, and combined with the system's preset swipe verification success rate model, the expected number of retries for the target user to complete a single swipe verification is calculated. The calculation uses a mapping function, substituting the swipe trajectory jitter characteristic value into the preset function model: Expected number of retries = Base number of retries + α × (Swipe trajectory jitter characteristic - Baseline jitter characteristic), where α is an adjustment coefficient, and the base number of retries and the baseline jitter characteristic are both preset parameters. The larger the user's swipe trajectory jitter characteristic value, the larger the calculated expected number of retries, and vice versa.

[0062] Step 1043: Multiply the total number of characters in the text input by the expected time to obtain the estimated input duration, and multiply the number of sliding verification nodes by the expected number of retries and the baseline time for a single sliding operation to obtain the estimated verification duration.

[0063] In this embodiment, the estimated input duration refers to the total time expected to be required for a target user to complete all text input operations in the application process, used to quantify the time cost of the text input stage. The total number of text input characters represents the total number of characters the user needs to manually input in the application process. The expected time consumption refers to the average time required for the user to input a single character. The estimated verification duration represents the total time expected to be required for the user to complete all sliding verification stages in the application process, reflecting the time cost of the security verification stage. The number of sliding verification nodes refers to the number of sliding verification stages that the user needs to complete in the application process. The expected number of retries represents the average number of attempts the user needs to make each time to complete a sliding verification. The baseline time for a single sliding verification operation refers to the standard time required to complete one sliding verification operation under ideal conditions, typically derived from statistics based on a large amount of user data.

[0064] Specifically, firstly, the total number of extracted text input characters is multiplied by the calculated expected time for inputting a single character to obtain the estimated input time. The formula is: Estimated Input Time = Total Number of Text Input Characters × Expected Time for Inputting a Single Character. For example, if the total number of text input characters is 50, and the expected time for inputting a single character is 0.8 seconds / character, then the estimated input time is 50 × 0.8 = 40 seconds. Next, the extracted number of sliding verification nodes is multiplied by the calculated expected number of retries and the system's preset baseline time for a single sliding step to obtain the estimated verification time. The formula is: Estimated Verification Time = Number of Sliding Verification Nodes × Expected Number of Retries × Baseline Time for a Single Sliding Step. For example, if there are 2 sliding verification nodes, the expected number of retries is 1.5, and the baseline time for a single sliding step is 3 seconds, then the estimated verification time is 2 × 1.5 × 3 = 9 seconds. These two calculations represent the estimated time a target user will spend completing the text input and sliding verification stages of the application process, respectively.

[0065] Step 1044: Sum the estimated input duration, estimated verification duration, and page loading time corresponding to the number of page transition steps to obtain the predicted total interaction time for the target user to complete the subscription process.

[0066] In this embodiment, the predicted total interaction time refers to the estimated total time required for a user to complete the entire subscription process, based on individual user characteristics and product interaction complexity. This estimate is used to assess the time cost for a specific user to operate a specific product. The estimated input time represents the total time expected for a user to complete all text input operations in the subscription process. The estimated verification time represents the total time expected for a user to complete all sliding verification steps in the subscription process. The page transition steps represent the number of page switches required to complete the entire subscription process. The page loading time refers to the time required for the system to load new page content during each page switch, including data transmission and interface rendering time. The page loading time corresponding to the page transition steps represents the cumulative time consumed by all page switches in the subscription process.

[0067] Specifically, the system first obtains the calculated estimated input time and estimated verification time, representing the estimated time the user will spend on text input and sliding verification, respectively. Then, based on the extracted page transition steps and the system's preset average page load time parameter, the cumulative loading time during the page transition is calculated. The formula is: Page Load Time = Page Transition Steps × Average Page Load Time. For example, if there are 6 page transition steps and the average page load time is 2 seconds, the page load time is 6 × 2 = 12 seconds. Finally, the above three times are summed to calculate the predicted total interaction time. The formula is: Predicted Total Interaction Time = Estimated Input Time + Estimated Verification Time + Page Load Time. For example, if the estimated input time is 40 seconds, the estimated verification time is 9 seconds, and the page load time is 12 seconds, the predicted total interaction time is 40 + 9 + 12 = 61 seconds. This value represents the total time the system predicts the target user will need to complete the entire subscription process for the candidate financial product.

[0068] Step 1045: Obtain the security session timeout duration of the subscription process for candidate financial products; when the predicted total interaction time exceeds the security session timeout duration, determine that the operational complexity of the candidate financial product exceeds the standard and remove the corresponding candidate financial product, retaining the remaining candidate financial products to form the initial screening product set.

[0069] In this embodiment, the secure session timeout duration refers to the maximum effective time of a single operation session set by the financial system for security reasons, used to prevent security risks caused by prolonged inactivity. Candidate financial products represent investment and wealth management products to be recommended to the target user after preliminary screening. Predicted total interaction time refers to the total time estimated by the system for a specific user to complete the product subscription process. Exceeding operational complexity limits indicates that the complexity of the product subscription process exceeds the user's acceptable range or system security restrictions. The initial product set refers to the set of financial products suitable for the target user's operational characteristics after operational complexity screening, serving as a candidate pool for subsequent personalized recommendations.

[0070] Specifically, the system first retrieves the security session timeout duration set in the subscription process of candidate financial products by querying their system configuration parameters. This security session timeout duration is typically pre-set by the financial product provider based on security policies and business characteristics, and is a key parameter for ensuring transaction security. Next, the calculated predicted total interaction time is compared with the retrieved security session timeout duration. The comparison uses a size judgment: if the predicted total interaction time exceeds the security session timeout duration, a judgment operation is performed, marking the candidate financial product as "operational complexity exceeds the limit." For all candidate financial products marked as "operational complexity exceeds the limit," the system removes them from the recommendation candidate list and no longer recommends them to the target user. Finally, all candidate financial products that were not removed are retained; the predicted total interaction time of these products does not exceed their security session timeout duration, matching the target user's operational capability characteristics. These retained candidate financial products constitute the initial screening product set, serving as input for subsequent personalized recommendation algorithms. For example, if a product's security session timeout is 60 seconds and the predicted total interaction time is 75 seconds, the product will be deemed to have exceeded the operational complexity limit and will be removed from the product list.

[0071] Step 105: Based on digital device operation proficiency and visual cognitive load, extract information and restructure the layout of the product description texts of each financial product in the initial screening product set to generate corresponding accessible display cards.

[0072] In this embodiment, the accessible display card refers to a financial product information display unit optimized based on the user's digital capabilities. Through information simplification, layout optimization, and adjustments to interactive elements, it reduces the cognitive burden on users in understanding and operation. Product description text represents detailed textual information describing the characteristics, terms, and risks of financial products, typically containing numerous technical terms and complex clauses. Information extraction and layout reconstruction refers to the process of extracting key information from the original product description and reorganizing the layout based on the user's capabilities, aiming to reduce information complexity and improve user comprehension efficiency.

[0073] Specifically, firstly, based on the calculated visual cognitive load value, the upper limit of the number of characters to be retained in the target text is determined. This is obtained through a corresponding mapping function; the higher the visual cognitive load, the lower the upper limit of the number of characters to be retained. Then, based on a pre-set financial product dictionary, string matching and pattern recognition are performed on the product description text of each financial product in the initial screening product set to extract target transaction data, including term values ​​(e.g., 1 year, 3 months), amount values ​​(e.g., minimum investment amount, maximum limit), and risk levels (e.g., R1, R2). Next, based on the determined upper limit of the number of characters to be retained in the target text, a text template of suitable length is retrieved from a pre-set age-appropriate expression template library. These templates are typically designed with concise and intuitive expressions. The extracted target transaction data is then filled into the selected text template to generate summary sentences. Simultaneously, based on the calculated device operation proficiency, the touch response area of ​​the voice broadcast trigger control is determined. This is obtained through a corresponding mapping function; the lower the device operation proficiency, the larger the touch response area, making it easier for users to click. Finally, the designed voice broadcast trigger control and the generated summary sentences are combined and arranged according to the preset layout rules, and packaged to generate the final accessible display card.

[0074] In one possible implementation, based on digital device proficiency and visual cognitive load, information is extracted and the layout is restructured from the product description texts of each financial product in the initial screening product set to generate corresponding accessible display cards. Specifically, this includes steps 1051-1053, as follows: Step 1051: Determine the upper limit of the number of characters to be retained in the target text based on visual cognitive load; perform string matching on the product description text based on the preset financial product dictionary, and extract the target transaction data, which includes the term, amount and risk level.

[0075] In this embodiment, target transaction data refers to the core transaction element information extracted from the financial product description text, representing the product's most basic characteristic description, including key parameters such as term, amount, and risk level. Visual cognitive load represents the degree of psychological effort required by a user when reading and processing visual information, reflecting the user's tolerance for information density and complexity. The maximum number of characters to retain in the target text refers to the maximum number of characters in the product description text determined based on the user's visual cognitive ability, used to control the complexity of information display. The financial product dictionary represents a pre-set structured vocabulary set containing financial terminology, transaction elements, and key parameters, used to assist in text parsing and information extraction. The product description text refers to the original textual material provided by the financial institution describing the product's characteristics, transaction conditions, and risk warnings, typically containing a large number of technical terms and complex expressions. String matching refers to the computational process of identifying and extracting specific information by comparing the similarity between character sequences in the text and pre-set vocabulary. Term represents the investment period of the financial product, usually in days, months, or years. Amount refers to the investment threshold or amount range of the financial product, reflecting the product's capital requirements. Risk rating is a graded assessment of the degree of risk of a financial product, usually expressed as R1-R5 or low, medium, and high levels.

[0076] Specifically, firstly, based on the user's visual cognitive load parameter, a mapping function is used to determine the upper limit of the target text's word count. The calculation employs a piecewise function: when the visual cognitive load index is less than threshold T1, the upper limit of the target text's word count is set to N1 (e.g., 50 words); when the visual cognitive load index is between thresholds T1 and T2, the upper limit is set to N2 (e.g., 100 words); and when the visual cognitive load index is greater than threshold T2, the upper limit is set to N3 (e.g., 150 words). After determining the upper limit of the target text's word count, a pre-defined financial product dictionary is loaded. This dictionary contains standard expressions and variations of key transaction elements such as term, amount, and risk level. Then, the product description text is segmented into word sequences. Next, a sliding window approach is used to perform string matching on each word in the text and its context, calculating the similarity with entries in the dictionary. When the similarity exceeds a pre-defined threshold (e.g., 0.8), the text segment is marked as a successful match, and the corresponding target transaction data is extracted based on the type of the matched dictionary entry. For term data, extract numerical representations including time units such as "days," "months," and "years"; for monetary data, extract representations including currency units or numbers plus "ten thousand" or "hundred million"; for risk levels, extract standard risk ratings including "R1" to "R5" or "low risk," "medium risk," and "high risk." Finally, organize all extracted information into a structured format, including target transaction data with three dimensions: term, amount, and risk level.

[0077] Step 1052: Based on the maximum number of characters to be retained in the target text, select a text template of the corresponding length from the preset age-appropriate expression template library, and fill the target transaction data into the text template to generate summary sentences.

[0078] In this embodiment, the summary sentence refers to a concise description of the core information of a financial product generated through a structured template, representing a summary of the product's key points that has been processed for age-friendliness to facilitate quick user understanding. The maximum number of characters in the target text represents the maximum number of characters in the text determined based on the user's visual cognitive load, used to control the complexity of the information display. The age-friendly expression template library refers to a collection of language expression templates specifically designed for elderly users, characterized by the use of simple and straightforward vocabulary, clear grammatical structures, and appropriate font sizes, making it easier for older adults to understand and accept. The text template represents a pre-defined text framework with a fixed structure and variable parameter positions, used to generate standardized product descriptions. The target transaction data refers to the core financial transaction elements extracted from the product description text, including key information such as term, amount, and risk level.

[0079] Specifically, firstly, based on the determined upper limit of the target text's character count, a selection process is conducted from the age-appropriate expression template library. The selection process is based on the standard character count of the templates, choosing those templates whose total character count after filling does not exceed the upper limit of the target text's character count. The templates in the age-appropriate expression template library are categorized by length into short sentence templates (under 30 characters), medium sentence templates (30-80 characters), and long sentence templates (over 80 characters). Information completeness is prioritized during selection; templates that can accommodate more target transaction data are chosen while still meeting the character count limit. After selecting a suitable text template, the target transaction data extracted in step 1051 is filled in according to the parameter positions in the template. The filling process uses a template string replacement method, replacing the parameter placeholders in the template (such as "{term}", "{amount}", "{risk level}") with the actual target transaction data values. For numeric data (such as amounts), a format conversion is performed, converting "10000" to more easily understood forms such as "10,000". For time-related data (such as time limits), standardize units to ensure the use of clear time units such as "days," "months," and "years." For risk levels, convert professional classifications (such as "R3") into more colloquial expressions (such as "medium risk"). After data population, validate the generated text to ensure all parameters are correctly filled in, and make necessary grammatical adjustments to generate summary sentences that conform to natural language expression habits. If the populated text exceeds the target text's character limit, try using a more concise alternative template or reducing non-core information to ensure the final generated summary sentences do not exceed the character limit.

[0080] Step 1053: Determine the touch response area of ​​the voice broadcast trigger control based on the user's proficiency in operating digital devices; combine and arrange the voice broadcast trigger control with the summary phrases, and encapsulate them to generate an accessible display card.

[0081] In this embodiment, the accessible display card refers to an information display unit optimized for specific user groups (such as the elderly and visually impaired individuals), representing an interactive interface component that combines visual and auditory channels to convey financial product information. Digital device operation proficiency indicates a user's level of ability to use digital terminals such as smartphones and tablets, typically derived through analysis of historical user operation data. The voice broadcast trigger control refers to an interactive element used to activate the voice reading function, allowing users to hear product information through simple operations such as clicking and touching. Touch response area indicates the actual area of ​​the control that can receive user touch operations, corresponding to the physical size of the interface element. The summary sentence refers to a description of the core information of the financial product after age-appropriate processing. The layout refers to the process of spatially arranging and styling text content and interactive controls according to visual design principles.

[0082] Specifically, firstly, based on the digital device operation proficiency parameters recorded in the user profile, the touch response area of ​​the voice broadcast trigger control is calculated using a mapping function. The calculation employs a piecewise function: when the digital device operation proficiency score is below L1 (low-proficiency user), the touch response area is set to 1.5 times the standard size, i.e., S1; when the digital device operation proficiency score is between L1 and L2 (medium-proficiency user), the touch response area is set to 1.2 times the standard size, i.e., S2; when the digital device operation proficiency score is above L2 (high-proficiency user), the touch response area is set to the standard size, i.e., S3. After determining the touch response area, a voice broadcast trigger control template of the corresponding size is selected, including the control icon, background area, and trigger hotspot. Then, the generated summary phrase is combined and formatted with the selected voice broadcast trigger control. The layout follows these rules: the text area occupies the main part of the display card, approximately 70%-80% of the total area; the voice prompt trigger control is located in the upper right corner or the center of the right side of the card, ensuring easy accessibility; the text uses an accessible font, with a font size no smaller than 1.2 times the minimum readable font size; the line spacing is set to 1.5 times the font size to ensure clear legibility; sufficient white space is maintained between the text and controls to avoid accidental touches. After completing the layout design, the overall visual optimization is performed, including setting an appropriate background color (usually choosing a high-contrast color scheme), adding borders to enhance boundary awareness, and applying rounded corners to improve user-friendliness. Finally, the text content, control elements, style information, and interaction logic are encapsulated into a unified accessible display card component. This component has responsive layout capabilities and can adapt to display devices with different screen sizes and resolutions.

[0083] Step 106: Calculate the visual information density of each accessibility display card, arrange the accessibility display cards according to the visual information density, generate and output a recommended list of age-friendly financial products.

[0084] In this embodiment, visual information density refers to the amount of information contained within a unit area of ​​the display card, representing the compactness of the information arrangement, and directly affecting the user's visual perception difficulty and information acquisition efficiency. The age-friendly financial product recommendation list represents a collection of financial products sorted with optimized information density. Its display order and layout are specifically designed for the visual cognitive characteristics of elderly users and other special groups to reduce the information acquisition threshold. A single-column waterfall layout refers to a vertically arranged content display method where each display card is arranged in a single column. Users can browse all content by vertically swiping, reducing visual interference and cognitive burden caused by complex multi-column layouts.

[0085] Specifically, firstly, the screen pixel area occupied by each generated accessibility display card and the total number of characters it contains are calculated. The pixel area is obtained by multiplying the card's height by its width, and the total number of characters is obtained by counting the number of text characters contained in the card. Then, the ratio of the total number of characters to the screen pixel area is calculated to obtain the visual information density of each accessibility display card, which represents the number of characters contained per unit area. Next, all accessibility display cards are sorted in ascending order of visual information density, with cards of lower information density appearing first and cards of higher information density appearing last. Finally, the sorted accessibility display cards are rendered onto the AI ​​terminal screen in a single-column waterfall layout, forming a vertically arranged product display list. Users can browse all recommended products through simple up and down scrolling, thereby generating and displaying a recommended list of age-friendly financial products.

[0086] In the above embodiments, basic accessible display cards for financial product information were generated by extracting target transaction data and using age-friendly expression templates. To further optimize the information acquisition experience for elderly users and reduce the impact of visual cognitive load on financial decision-making, this application also provides an artificial intelligence-based financial product recommendation method. This method calculates the visual information density of the display cards in real time to achieve orderly arrangement, constructs a reading state perception mechanism based on user touch behavior, and combines the text decoding ability of elderly users to adaptively transform the content format, enabling the system to more humanely and accurately meet the cognitive load management and multimodal assistance needs of elderly users in the process of acquiring complex financial information. The following section will combine... Figure 2 Another method for recommending financial products based on artificial intelligence, as described in the embodiments of this application, is as follows: Please see Figure 2 This is a flowchart illustrating an artificial intelligence-based financial product recommendation method in an embodiment of this application.

[0087] Step 201: Calculate the screen pixel area occupied by each accessibility display card and the total number of characters contained therein.

[0088] In this embodiment, the accessible display card refers to an information display unit specifically designed for visually impaired or elderly users, featuring appropriate font size, contrast, and assistive functions to facilitate the perception and understanding of financial product information by this specific group. Screen pixel area refers to the actual number of pixels occupied by the accessible display card on the terminal display screen, typically measured in square pixels, reflecting the visual space occupied by the card. Total character count represents the number of characters in all visible text content contained within the accessible display card, including punctuation marks, spaces, and special symbols, used to quantify the information capacity within the card. Statistics refers to the process of obtaining the area and character count parameters of the accessible display card through calculation, providing a data foundation for subsequent calculations of visual information density.

[0089] Specifically, the process begins by retrieving a collection of all rendered accessibility display card elements on the current page. For each card element, its position and size information on the screen are obtained through DOM manipulation, including the coordinates of its top-left and bottom-right corners. Based on these coordinates, the card's width and height are calculated, and then the screen pixel area occupied by the card is calculated by multiplying the width by the height, with the result in square pixels. Simultaneously, all text node content within the card is extracted, including titles, descriptions, and data tags, and concatenated into a complete text string. The length of the concatenated string is calculated to obtain the total number of characters. For cards containing non-text elements such as icons and buttons, only the characters in the text portion are counted, excluding the space occupied by non-text elements. For multilingual text, Unicode code points are used for counting, with each code point counting as one character. The calculated screen pixel area and total number of characters are stored in the properties of the corresponding card object for subsequent visual information density calculations. This process is repeated until all accessibility display cards are statistically analyzed.

[0090] Step 202: Calculate the ratio of the total number of characters to the screen pixel area to obtain the visual information density of each accessibility display card.

[0091] In this embodiment, visual information density refers to the number of characters contained within a unit screen area, representing the compactness or sparseness of information in interface elements, and is typically measured in units of characters per unit area. Total characters represent the total number of visible text characters contained in the accessibility display card. Screen pixel area refers to the total number of pixels occupied by the card on the display screen. Ratio calculation refers to obtaining the character distribution density per unit area through division. Accessibility display cards represent information display components optimized for specific user groups, and their visual information density directly relates to the user's reading experience and information acquisition efficiency.

[0092] Specifically, for each accessible display card, two parameters are extracted: the total number of characters and the screen pixel area. For each card, its visual information density is calculated by dividing the total number of characters by the screen pixel area, with the result expressed as the number of characters per square pixel. Since the pixel area is usually a large value while the number of characters is relatively small, the calculated visual information density is typically a very small decimal. To facilitate numerical processing and comparison, the calculation result can be scaled, such as by multiplying it by a large constant (e.g., millions) to obtain a more manageable value, namely, the number of characters per million pixels. During processing, if the screen pixel area of ​​a card is zero (which should theoretically not occur but is treated as an edge case), the visual information density of that card is set to a maximum value to ensure it is identified as an outlier in subsequent sorting. For cards with a total number of characters of zero (such as pure image cards), the calculated visual information density is zero. After calculation, a visual information density attribute is associated with each accessible display card for subsequent sorting operations.

[0093] Step 203: Arrange the accessibility display cards in ascending order according to visual information density; render the ascending-ordered accessibility display cards onto the screen of the AI ​​terminal in a single-column waterfall layout to generate and display a list of recommended age-friendly financial products.

[0094] In this embodiment, the "age-friendly financial product recommendation list" refers to a financial product information display interface specifically designed for elderly users. It represents a collection of content designed to enable elderly users to easily access, understand, and operate financial product information through optimized visual presentation, simplified interaction logic, and adjusted information density. Visual information density refers to the number of characters contained within a unit screen area, reflecting the compactness of interface elements. Ascending order refers to the process of reordering accessibility display cards according to their visual information density from smallest to largest. Single-column waterfall layout refers to a vertically arranged interface layout where content elements are arranged sequentially in a single column, allowing users to scroll vertically to view all content, suitable for linear reading habits. Accessibility display cards represent information display units optimized for specific user groups. Artificial intelligence terminals refer to electronic devices with intelligent interaction and adaptive interface capabilities, such as smartphones and tablets. Rendering refers to the technical process of converting data into visual interface elements and presenting them on the screen.

[0095] Specifically, the process begins by obtaining a set of all calculated accessibility display cards and their visual information density values. These cards are then sorted in ascending order of visual information density, using a comparison function to ensure that cards with lower visual information density appear first. A stable sorting algorithm is employed to ensure that cards with the same visual information density maintain their original order. After sorting, an empty container element is created as the outer container for the waterfall layout, with its width set to a majority of the screen width (e.g., 95%) and centered. Each accessibility display card element is then added to the container in the sorted order. For each card element, its width is set to the full width of the container, filling the entire container space horizontally to create a single-column layout. Fixed spacing is added between adjacent cards, typically set to an appropriate pixel value, to create clear visual separation. The container is given a vertical scrollable property, allowing vertical scrolling when the content exceeds the screen height. To improve scrolling smoothness, a virtual scrolling mechanism can be enabled for the container, rendering only cards within the currently visible viewport area and the pre-loaded areas before and after it. A uniform animation transition effect, such as fade-in or swipe-up, is applied to the arranged card set to enhance the interactive experience. Finally, the completed container is rendered to the designated display area of ​​the AI ​​terminal to form a complete list of recommended age-friendly financial products, and the screen is refreshed to be displayed to the user.

[0096] In one possible implementation, after generating and displaying a list of recommended financial products suitable for the elderly, the process further includes steps 2031-2034, as follows: Step 2031: Collect the continuous touch coordinates of the target user on the recommended list of age-friendly financial products page.

[0097] Specifically, the touch event listener is first initialized and bound to the root container element of the age-friendly financial product recommendation list page. When a touch start event is detected, a new empty coordinate point array is created to store all coordinate points during this continuous touch process. The initial position coordinates and timestamp at the start of the touch are recorded as the first touch coordinate point. During the user's continuous screen touch, the movement trajectory of the finger on the screen is captured in real time through touch movement events. For each touch movement event, the current touch position coordinates (x, y) and the corresponding timestamp t are recorded to form a triple (x, y, t), and added to the coordinate point array. To avoid data redundancy, a reasonable sampling frequency is set, such as sampling coordinate points every 50 milliseconds, or recording a new coordinate point only when the distance between two sampling points exceeds a certain threshold (e.g., 5 pixels). At the same time, noise data that may be caused by sensor jitter is filtered out, such as only keeping one point when two adjacent sampling points are too close (less than 2 pixels). For multi-touch, only the coordinate data of the first touch point is recorded, ignoring other touch points. When a touch end event or touch cancel event is detected, the acquisition process of the current continuous touch coordinates is completed, and the complete coordinate point array is passed to subsequent steps for analysis and processing. The entire acquisition process maintains a high sampling rate (at least 60Hz) to ensure the capture of fine finger micro-movement features.

[0098] Step 2032: When the continuous touch coordinate points fall within the display area of ​​the target accessible display card for a preset duration, and the trajectory dispersion radius formed by the continuous touch coordinate points is less than the maximum offset corresponding to the sliding trajectory jitter feature, it is determined that the target user has entered the finger anchoring reading state.

[0099] Specifically, firstly, a subset of touch points within the most recent period (e.g., the past 3 seconds) is extracted from the obtained array of continuous touch coordinates. It is then determined whether all these touch points fall within the same accessible display card's display area. If any touch point falls outside the card's area, the judgment process is reset. The time span of these touch points is calculated, i.e., the timestamp of the last touch point minus the timestamp of the first touch point, to obtain the touch duration. The touch duration is compared with a preset duration (e.g., 1.5 seconds). Only if the duration is greater than or equal to the preset duration does the subsequent judgment continue. The geometric center point of all touch points is calculated by averaging all x and y coordinates. The distance from each touch point to the geometric center point is calculated, and the maximum distance value is used as the trajectory dispersion radius. The user's swipe trajectory jitter characteristics are obtained from the user profile data, and the corresponding maximum offset parameter (e.g., 15 pixels) is extracted. The calculated trajectory dispersion radius is compared with the maximum offset. If the dispersion radius is less than the maximum offset, it is considered that the user's finger remains relatively still rather than intentionally swiping. When both conditions are met simultaneously—"the touch point remains within the same card area for a preset duration" and "the trajectory dispersion radius is less than the maximum offset"—the user is determined to have entered the finger-anchored reading state, and the start timestamp of this state is recorded. To avoid frequent state switching, a state lockout period can be set, so that the state is not re-evaluated for a short period of time (e.g., 0.5 seconds) after entering the anchored state.

[0100] Step 2033: Calculate the duration of finger-anchored reading state and compare the duration with the information acquisition time cost corresponding to visual cognitive load.

[0101] Specifically, the system obtains the start timestamp of when the user enters the finger-anchored reading state. It monitors the user's touch state in real time, and when the user's touch behavior no longer meets the criteria for the finger-anchored reading state (e.g., finger movement exceeds the maximum offset, finger is lifted off the screen, touch point moves out of the current card area), it records the timestamp of that moment as the end timestamp. The duration of the finger-anchored reading state is calculated by subtracting the start timestamp from the end timestamp; the difference is the duration in milliseconds. The system retrieves the current user's visual cognitive load parameter value from the user profile database. This parameter is typically derived based on a comprehensive evaluation of the user's age, cognitive ability test results, and historical interaction data. Based on the visual cognitive load parameter and the amount of information (e.g., number of characters, information complexity) in the target accessible display card, it calculates the theoretical information acquisition time cost required for the user to understand the card content. The calculation method typically uses a linear model: Information acquisition time cost = Basic reading time + Visual cognitive load coefficient × Information volume. The basic reading time is the average time required for a standard user to read the same content (e.g., 200 characters per minute); the visual cognitive load coefficient is a time extension coefficient adjusted according to the user's visual cognitive load, with a larger coefficient for higher loads; and the information content is the number of effective information units contained in the card. The calculated duration of the finger-anchored reading state is numerically compared with the information acquisition time cost to determine their relative magnitudes. The comparison results are recorded for subsequent processing decisions. To improve the accuracy of the comparison, a weighted average of historical comparison results can be considered to reduce the impact of random fluctuations in a single measurement.

[0102] Step 2034: When the duration exceeds the information acquisition time cost, it is determined that the target user has a text decoding lag in the target accessibility display card; the summary short sentences in the target accessibility display card are folded and replaced with the visualization graphics corresponding to the target transaction data, and the voice parsing and broadcasting of the target transaction data are activated simultaneously.

[0103] Specifically, the process begins by comparing the duration of the finger-anchored reading state with the time cost of information retrieval. When the duration exceeds the time cost of information retrieval, and the excess exceeds a certain threshold (e.g., over 20%), it is determined that the user is experiencing text decoding lag. For cases where text decoding lag is identified, the DOM element reference of the target accessibility display card is obtained, and the text container element containing the summary phrase is located. The summary phrase text container is then collapsed and hidden by modifying CSS properties (e.g., setting display:none or height:0 with overflow:hidden). Based on the target transaction data in the card (term, amount, risk level, etc.), a suitable visualization type is determined: for term data, a simplified calendar or clock graphic can be used; for amount data, currency symbol graphics of varying heights can be used; for risk level, color-coded risk level icons can be used. Corresponding visualization DOM elements are created, with appropriate size, color, and animation effects set to ensure visual appeal meets age-friendly design standards, such as using high-contrast color schemes, simple graphics, and sufficiently large sizes. The created visualization elements are then inserted into the original summary phrase position, replacing the text content. Simultaneously, the voice broadcast content is constructed, formatting the target transaction data into concise and easy-to-understand language descriptions, such as "The product term is 3 months, the minimum investment amount is 10,000 yuan, and the risk level is medium." The system's speech synthesis engine is invoked to play the constructed voice content at a moderate speaking speed (e.g., 180 words per minute) and a volume suitable for the elderly (e.g., 1.2 times the system default volume). During the voice playback, corresponding visual graphics can be given synchronized emphasis effects; for example, when a certain data point is broadcast, the corresponding graphic briefly enlarges or highlights. To avoid interrupting the user experience, a small control button is added, allowing users to manually stop the voice broadcast. This auxiliary interaction event is recorded for subsequent user experience optimization and personalized adjustments.

[0104] The following describes an AI-based financial product recommendation system from a hardware processing perspective, as described in the embodiments of this invention. Please refer to [link / reference]. Figure 3 This is a schematic diagram of the structure of an artificial intelligence-based financial product recommendation system in an embodiment of this application.

[0105] It should be noted that, Figure 3 The structure of the AI-based financial product recommendation system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0106] like Figure 3As shown, an artificial intelligence-based financial product recommendation system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 302 or programs loaded from storage portion 308 into Random Access Memory (RAM) 303, such as executing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0107] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0108] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0109] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0110] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0111] Specifically, an AI-based financial product recommendation system according to this embodiment includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it implements the AI-based financial product recommendation method provided in the above embodiment.

[0112] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the AI-based financial product recommendation system described in the above embodiments; or it may exist independently and not incorporated into the AI-based financial product recommendation system. The storage medium carries one or more computer programs, which, when executed by a processor of the AI-based financial product recommendation system, enable the AI-based financial product recommendation system to implement the AI-based financial product recommendation method based on encrypted data transmission via the Internet of Things provided in the above embodiments.

Claims

1. A method for recommending financial products based on artificial intelligence, characterized in that, The method includes: Acquire data on the physical interaction behavior of target users on AI terminals and the underlying configuration parameters of the terminal system; Extract the touch dwell time and swipe trajectory jitter features from the physical interaction behavior data, as well as the display font scaling ratio from the underlying configuration parameters; The device operation proficiency and visual cognitive load of the target user are determined based on the touch dwell time, the swipe trajectory jitter characteristics, and the display font scaling ratio. Extract the operational complexity tags of candidate financial products, match and compare the operational complexity tags with the device operation proficiency, and eliminate candidate financial products whose device operation complexity exceeds the standard to obtain the initial screening product set; Based on the proficiency in operating the digital devices and the visual cognitive load, information is extracted and the layout is reconstructed from the product description texts of each financial product in the initial screening product set to generate corresponding accessible display cards. Calculate the visual information density of each of the accessible display cards, arrange the accessible display cards according to the visual information density, generate a recommended list of age-friendly financial products, and output it.

2. The method according to claim 1, characterized in that, The extraction of touch dwell time and swipe trajectory jitter features from the physical interaction behavior data includes: The touchscreen sensor interface of the AI ​​terminal is invoked to obtain the time difference between when the target user's finger touches the screen and when it leaves the screen during a single touch operation, which is used as the touch dwell time. When the target user performs a swipe operation, the touch screen sensor interface collects a continuous set of screen coordinate points at a preset sampling frequency. Based on the continuous set of screen coordinate points, calculate the sliding direction angle formed by the line connecting two adjacent screen coordinate points; Calculate the angle difference between adjacent sliding direction angles, and use the statistical variance of the angle difference as the sliding trajectory jitter feature.

3. The method according to claim 1, characterized in that, The step of determining the target user's device operation proficiency and visual cognitive load based on the touch dwell time, the swipe trajectory jitter characteristics, and the display font scaling ratio includes: The ratio of the touch dwell time to a preset baseline dwell time is calculated as a sluggishness coefficient, and the ratio of the sliding trajectory jitter characteristics to a preset baseline jitter variance is calculated as a vibration coefficient. The operating resistance index is obtained by weighting the sluggishness coefficient and the flutter coefficient, and the operating proficiency of the digital device is obtained by dividing the preset benchmark proficiency by the operating resistance index. Calculate the single-screen character capacity of the current smart terminal screen window based on the display font scaling ratio; Calculate the ratio of the preset number of characters in the standard terms of financial products to the character capacity of a single screen to obtain the expected number of page turns; Multiplying the expected number of page turns by the touch duration yields the visual cognitive load, which represents the time cost of information acquisition.

4. The method according to claim 1, characterized in that, The process involves extracting operational complexity tags from candidate financial products, matching these tags with the device's operational proficiency level, and eliminating candidate financial products whose operational complexity exceeds the threshold, resulting in a preliminary product set, including: The interaction nodes of the candidate financial products in the subscription process are analyzed, and the total number of text input characters, the number of sliding verification nodes, and the number of page transitions are extracted based on the interaction nodes, and combined to form the operation complexity label. Based on the target user's touch dwell time and the swipe trajectory jitter characteristics, calculate the expected time for the target user to complete a single character input and the expected number of retries to complete a single swipe verification. Multiply the total number of characters in the text input by the expected time to obtain the estimated input time, and multiply the number of sliding verification nodes by the expected number of retries and the baseline time for a single sliding operation to obtain the estimated verification time. The predicted total interaction time for the target user to complete the subscription process is obtained by summing the estimated input time, the estimated verification time, and the page loading time corresponding to the number of page transition steps. Obtain the secure session timeout duration for the subscription process of the candidate financial products; When the predicted total interaction time exceeds the secure session timeout time, the operational complexity of the candidate financial product is determined to be excessive, and the corresponding candidate financial product is removed, while the remaining candidate financial products constitute the initial screening product set.

5. The method according to claim 1, characterized in that, Based on the user's proficiency in operating the digital device and the visual cognitive load, information is extracted and the layout is restructured from the product description texts of each financial product in the initial screening product set to generate corresponding accessible display cards, including: Based on the aforementioned visual cognitive load, determine the upper limit of the number of characters to be retained in the target text; Based on a pre-set financial product dictionary, string matching is performed on the product description text to extract target transaction data, which includes term, amount, and risk level. Based on the maximum number of characters to be retained in the target text, a text template of the corresponding length is selected from the preset age-appropriate expression template library, and the target transaction data is filled into the text template to generate a summary sentence; The touch response area of ​​the voice broadcast trigger control is determined based on the user's proficiency in operating the digital device. The voice broadcast trigger control and the summary phrase are combined and arranged to generate the accessibility display card.

6. The method according to claim 1, characterized in that, The process of calculating the visual information density of each of the accessible display cards, arranging the accessible display cards according to the visual information density, generating and outputting a recommended list of age-friendly financial products includes: Calculate the screen pixel area occupied by each of the aforementioned accessibility display cards and the total number of characters contained therein; Calculate the ratio of the total number of characters to the screen pixel area to obtain the visual information density of each of the accessibility display cards; The accessibility display cards are arranged in ascending order according to the visual information density. The accessibility display cards, arranged in ascending order, are rendered onto the screen of the AI ​​terminal in a single-column waterfall layout, generating and displaying the recommended list of age-friendly financial products.

7. The method according to claim 6, characterized in that, After generating and displaying the recommended list of age-friendly financial products, the process also includes: Collect the continuous touch coordinates of the target user on the recommended list page of age-friendly financial products; When the continuous touch coordinate points fall within the display area of ​​the target accessibility display card for a preset duration, and the trajectory dispersion radius formed by the continuous touch coordinate points is less than the maximum offset corresponding to the sliding trajectory jitter feature, it is determined that the target user has entered the finger anchoring reading state. The duration of the finger-anchored reading state is statistically analyzed, and the duration is compared with the information acquisition time cost corresponding to the visual cognitive load. When the duration exceeds the information acquisition time cost, it is determined that the target user has a text decoding lag in the target accessibility display card; The summary phrases in the target accessibility display card are folded and replaced with visual graphics corresponding to the target transaction data, while simultaneously activating voice analysis and broadcasting for the target transaction data.

8. A financial product recommendation system based on artificial intelligence, characterized in that, The AI-based financial product recommendation system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors invoke the computer instructions to cause the AI-based financial product recommendation system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the AI-based financial product recommendation system, the AI-based financial product recommendation system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the AI-based financial product recommendation system, the AI-based financial product recommendation system performs the method as described in any one of claims 1-7.

Images