A community elderly population intelligent health assessment management system and method

Abstract

The application discloses a kind of community old population intelligent health assessment management system and method, it is related to wisdom medical treatment and health data processing technical field.The system includes user management, five-dimensional health assessment, breakpoint continuation evaluation and synchronization, intelligent data analysis, individualized recommendation and old adaptation interaction etc.Modulus.Its method covers to establish encrypted health record;When obtaining multidimensional original evaluation data, collect answer behavior index, generate correction coefficient to carry out dynamic multiplier calibration to score;The latest data state is locked to realize breakpoint continuation evaluation recovery by two-way time stamp checking mechanism;Health score time series fitting model is constructed, and risk signal is generated based on multi-level early warning condition;Using weighted collaborative filtering matching algorithm, individualized intervention suggestion is generated by comprehensively indexing and record information;Finally, output is carried out in old adaptation visual rendering format.The application can dynamically calibrate evaluation result, improve the reliability of data recovery, and output high matching degree health intervention suggestion.

Description

Technical Field

[0001] This invention relates to the field of smart healthcare and health data processing technology, and in particular to a smart health assessment and management system and method for elderly people in communities. Background Technology

[0002] With the advancement of societal aging, the demand for health management among the elderly in communities is increasing. Regular health assessments are a crucial step in understanding the physical and mental health of the elderly, preventing chronic diseases, and developing intervention measures. Current health assessments typically rely on information platforms, collecting questionnaire data on various dimensions such as physical function and daily living abilities to reflect their current health level in the form of quantitative scores.

[0003] In actual assessments, older adults, limited by their physical functions and cognitive state, often take excessively long times or exhibit logical inconsistencies when answering questions. Conventional statistical methods frequently use raw response data directly for score conversion, failing to adequately consider the impact of behavioral characteristics on the objectivity of the final score. Furthermore, when older adults use mobile devices to participate in lengthy assessment tasks, assessments are easily interrupted due to misoperation or network fluctuations, often requiring them to re-answer questions or encountering inconsistencies in data versions across different devices upon reconnection.

[0004] On the other hand, existing systems, after generating assessment reports, typically provide health guidance based on a single-dimensional static score threshold for general matching, rarely considering an individual's past medical history, historical intervention feedback records, and longitudinal time-series health evolution trends. The elderly population has relatively lower sensitivity to information reception, and conventional visual interfaces and information interaction methods are difficult to fully adapt to the visual and auditory habits of elderly users in terms of font size, color contrast, and risk warnings. These objective shortcomings limit the application value of existing health assessment and management systems among the elderly population in the community. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned defects of the prior art and provide an intelligent health assessment management system and method for the elderly in the community. This invention addresses the problems in existing elderly health assessments, such as inaccurate assessments due to the lack of integration with behavioral characteristics of answering questions, easy data loss after interruption, and the lack of accurate and personalized matching and age-appropriate presentation of subsequent intervention suggestions. In this way, it can achieve objective quantitative assessment of the health status of the elderly, accurate risk warning, and age-appropriate intervention guidance.

[0006] In a first aspect, the present invention provides an intelligent health assessment and management system for elderly people in a community, the system comprising: The user management module is used to create health records for target users and store them in encrypted form. The five-dimensional health assessment module is connected to the user management module and is used to provide the target user with a multi-dimensional quantitative assessment task to obtain raw assessment data. During the execution of the assessment task, the module collects the target user's answer behavior indicators in real time, uses the answer behavior indicators to generate correction coefficients, and performs dynamic multiplier calibration on the preliminary health score calculated based on the raw assessment data to output the target health score. The breakpoint recovery and data synchronization module is connected to the five-dimensional health assessment module. It is used to save the original assessment data to the local cache in real time in the form of fragments and perform incremental cloud backup. When an abnormal assessment interruption is detected and the connection is restored, the latest data status is locked through the two-way timestamp verification mechanism between the local and cloud and the breakpoint recovery operation is triggered. The intelligent data analysis module is connected to the five-dimensional health assessment module and the breakpoint continuation assessment and data synchronization module, respectively. It is used to construct a time series fitting model based on the target health score of the target user in previous assessments, and generate a risk warning signal with the corresponding danger level when it is determined that the target health score or its time series change characteristics meet the multi-level warning triggering conditions. The personalized recommendation engine module is communicatively connected to the intelligent data analysis module and the user management module, respectively, and is used to obtain the target health score, the risk warning signal, the health record and historical suggestion feedback records. It uses a multi-dimensional weighted collaborative filtering matching algorithm to calculate the comprehensive matching degree of each candidate suggestion in order to generate personalized intervention suggestions for the target user. The age-friendly interactive interface module is communicatively connected to the personalized recommendation engine module and the five-dimensional health assessment module, and is used to output the personalized intervention suggestions and the risk warning signals in a visual rendering format that exceeds the preset benchmark parameters.

[0007] Optionally, when the five-dimensional health assessment module calculates a preliminary health score based on the original assessment data, it is specifically used for: The raw scores of the five dimensions of physical health, daily living ability, psychological state, cognitive function and social function included in the multi-dimensional quantitative assessment task are subtracted from the theoretical minimum extreme value of the corresponding dimension. The scores are then divided by the range of the corresponding dimension (i.e. the difference between the maximum score extreme value and the theoretical minimum extreme value). The calculation results are then mapped to a percentage benchmark range to obtain the standard score of each dimension. Obtain the weight coefficients that are independently configured for each of the above dimensions, multiply the standard score of each dimension by its corresponding weight coefficient, and sum all the product results to obtain the preliminary health score.

[0008] Optionally, the answering behavior indicators include time spent answering a single question, overall missed question rate, and logical consistency of answers; when the five-dimensional health assessment module uses the answering behavior indicators to generate correction coefficients for dynamic multiplier calibration, it is specifically used for: When at least one of the following abnormal situations is detected: the time taken to answer a single question is higher than a preset time threshold, the overall missing question rate exceeds a preset proportion, or the logical consistency of the answer is lower than a preset benchmark, it is determined that the current answering process is distorted due to fluctuations in the cognitive state of the target user. Based on the corresponding triggered abnormal situation, a preset penalty decay parameter is retrieved as the correction coefficient. The correction coefficient is multiplied and converted with the preliminary health score to output the target health score after dimensionality reduction adjustment.

[0009] Optionally, when the breakpoint recovery and data synchronization module locks the latest data status and triggers the breakpoint recovery operation through a two-way timestamp verification mechanism, it is specifically used for: The multi-dimensional quantitative evaluation task is divided into independent evaluation units. When the target user completes each evaluation item, the fragment data and progress status identifier in the local cache are updated immediately, and the incremental cloud backup is executed using an asynchronous thread. After the abnormal interruption is recovered, the first latest modified timestamp in the local cache and the second latest modified timestamp in the cloud database are extracted respectively. The first latest modified timestamp is compared with the second latest modified timestamp. The one with the closest time to the present is used as the reference data end. The data status in the reference data end is overwritten to the other end, and the age-friendly interactive interface module is guided to jump to the first incomplete evaluation item.

[0010] Optionally, when the intelligent data analysis module constructs a time series fitting model based on the target user's target health scores from previous assessments, it is specifically used for: Extract the target user's assessment data at different time points from the health record, and filter and remove extreme scores caused by abnormal operation. Linear interpolation algorithms are used to smoothly fill in missing time point data caused by evaluation interruption or omission, so as to generate a continuous health time series. The health time series is fitted using a linear regression algorithm to generate trend feature data that reflects the health evolution trajectory in each dimension.

[0011] Optionally, the multi-level early warning triggering conditions include static absolute value conditions, sequential cumulative decline conditions, and single sudden drop conditions; the risk early warning signals are divided into first-level early warning, second-level early warning, and third-level early warning. The intelligent data analysis module is configured to generate risk warning signals as follows: When the target health score or the score of any dimension in a single assessment falls into a predefined static low score range, the risk warning signal corresponding to the danger level is triggered based on the static absolute value condition. When the score of any dimension shows a continuous decline in consecutive evaluation batches and the cumulative decline reaches three increasing extreme value thresholds in sequence, the first level warning, the second level warning, or the third level warning is triggered based on the cumulative decline condition of the sequence. When it is determined that the difference between the single decrease in the score of any dimension in two adjacent assessments exceeds the preset sudden drop threshold, the static absolute value condition is masked, and the risk warning signal of no less than the second level warning is directly triggered based on the single sudden drop condition.

[0012] Optionally, when the personalized recommendation engine module calculates the comprehensive matching degree of each candidate suggestion using a multi-dimensional weighted collaborative filtering matching algorithm, it is specifically used for: For each candidate suggestion in the preset suggestion library, the matching degree of the evaluation result is calculated based on its health dimension label and the weakness dimension in the target health score; The user profile matching degree is calculated based on the applicable population tags of the candidate suggestions and the past medical history and age characteristics in the health records. The historical feedback matching degree is calculated based on the target user's effective marking of past intervention suggestions in the historical suggestion feedback record. Independent feature weight multipliers are assigned to the matching degree of the evaluation result, the matching degree of the user profile, and the matching degree of the historical feedback. The three are multiplied by their corresponding feature weight multipliers and then linearly superimposed to obtain the comprehensive matching degree of each candidate suggestion.

[0013] Optionally, the personalized recommendation engine module is further used for: All candidate suggestions with the calculated comprehensive matching degree are sorted in descending order; In response to the risk warning signal triggered by the intelligent data analysis module, the candidate suggestions that match the abnormal dimension that triggered the warning are forced to the top and their rank is promoted in the sorting sequence. After rank reconstruction, the personalized intervention suggestions are output.

[0014] Optionally, the age-friendly interactive interface module is configured to output the risk warning signal in a visual rendering format that exceeds preset baseline parameters as follows: The interface rendering employs a larger font structure than the conventional baseline threshold and a high-contrast color scheme. In response to the risk warning signals of different hazard levels, different levels of alert coloring channels are invoked in the visualization chart to highlight and dynamically annotate the abnormal dimensions, and a progressive intervention guidance voice broadcast stream is launched simultaneously.

[0015] Secondly, the present invention provides a method for intelligent health assessment and management of elderly people in communities, the method comprising the following steps: Step S1: Establish and encrypt the health records of the target users; Step S2: Provide the target user with a multi-dimensional quantitative assessment task to obtain raw assessment data, and collect the target user's answer behavior indicators in real time during the execution of the assessment task. Use the answer behavior indicators to generate correction coefficients to dynamically multiply and calibrate the preliminary health score calculated based on the raw assessment data, and output the target health score. Step S3: Save the original evaluation data to the local cache in real time in the form of fragments and perform incremental cloud backup, so that when an evaluation abnormal interruption is detected and reconnected, the latest data status is locked through the two-way timestamp verification mechanism between the local and cloud and the breakpoint recovery operation is triggered. Step S4: Construct a time series fitting model based on the target user's target health score in previous assessments, and generate a risk warning signal of the corresponding danger level when the target health score or its time series change characteristics meet the multi-level warning triggering conditions; Step S5: Obtain the target health score, the risk warning signal, the health record, and the historical suggestion feedback record; use a multi-dimensional weighted collaborative filtering matching algorithm to calculate the comprehensive matching degree of each candidate suggestion to generate personalized intervention suggestions for the target user. Step S6: Output the personalized intervention suggestion and the risk warning signal in a visual rendering format that exceeds the preset baseline parameters.

[0016] The present invention has achieved the following beneficial effects: This invention generates correction coefficients by collecting behavioral indicators such as single-question answering time, overall missed question rate, and logical consistency of answers during the answering process. These coefficients dynamically multiply and calibrate the preliminary health score calculated from the original data, improving the objectivity and accuracy of the final health assessment results. It introduces local caching and incremental cloud backup mechanisms, using bidirectional timestamp verification to lock the latest data status, ensuring consistency of data across multiple devices and a smooth review experience after assessment interruption. By constructing a time-series fitting model to analyze the health evolution trajectory, and combining static conditions with dynamic conditions such as sudden drops to trigger multi-level early warning signals, it improves the timeliness of risk detection and intervention. Using a multi-dimensional weighted collaborative filtering matching algorithm, it comprehensively considers the target health score, user profile, and historical feedback records to generate personalized intervention suggestions. The output uses an age-friendly visual rendering format, including enlarged font structure, high-contrast color scheme, and voice broadcasting, improving the effectiveness of information delivery and the system interaction experience for elderly users.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the internal structure of the community-based intelligent health assessment and management system for the elderly, as described in an embodiment of the present invention. Figure 2 This is a flowchart illustrating the intelligent health assessment and management method for the elderly population in the community, as described in this embodiment of the invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] It is understood that this application provides a smart health assessment and management system and method for elderly people in a community. This system uses an application or mini-program as the front-end interaction carrier, and the back-end is built on a cloud development server to achieve data storage and computation. The hardware environment on which this system relies covers various mobile terminal devices running standard operating systems. For example... Figure 1As shown, the system specifically includes: a user management module, a five-dimensional health assessment module, a breakpoint-based assessment and data synchronization module, an intelligent data analysis module, a personalized recommendation engine module, and an age-friendly interactive interface module. The technical details and working principles of each module are explained in detail below.

[0022] Specifically, the user management module is used to establish and encrypt the health records of target users. Upon first use, a target user initiates an access request through an authorized login interface or a guest mode interface. The user management module is configured with an authentication unit that generates a unique user ID upon receiving the access request. After the target user completes authentication, the user management module allocates dedicated storage space for that user in the cloud database and provides a data entry interface to obtain the target user's basic health data. This basic health data includes, but is not limited to, basic personal information, past medical history, daily medication use, physical examination records, and lifestyle characteristics.

[0023] Furthermore, to ensure data security, the user management module embeds an encryption processing engine. Before uploading the basic health data to the cloud database, the encryption processing engine calls advanced encryption standards (such as the AES-256 algorithm) to encrypt the target user's health record data. The specific encryption process is as follows: the system uses a pseudo-random number generator to generate a 256-bit symmetric encryption key, and groups the plaintext health record data into 128-bit data blocks; then, for each data block, it sequentially performs multiple rounds of logical and substitution operations, including byte substitution (SubBytes), row shifting (ShiftRows), column mixing (MixColumns), and round key addition (AddRoundKey), finally outputting a ciphertext data sequence. This ciphertext data sequence, along with data verification header information, is transmitted to the cloud database for persistent storage via an HTTPS encrypted secure channel.

[0024] Specifically, the five-dimensional health assessment module is communicatively connected to the user management module. It is used to provide the target user with a multi-dimensional quantitative assessment task to obtain raw assessment data. During the execution of the assessment task, it collects the target user's answer behavior indicators in real time, uses the answer behavior indicators to generate correction coefficients, and performs dynamic multiplier calibration on the preliminary health score calculated based on the raw assessment data to output the target health score.

[0025] The five-dimensional health assessment module internally constructs a quantitative assessment model, which includes a subset of items from a question bank covering five assessment dimensions: physical health, daily living abilities, psychological state, cognitive function, and social function. The multi-dimensional quantitative assessment task is parsed into an array of specific assessment items corresponding to the aforementioned five dimensions. The age-friendly interactive interface module sequentially calls and renders the array of assessment items according to a preset logical order, receives touch input commands from the target user, and then obtains the raw assessment data.

[0026] Furthermore, when the five-dimensional health assessment module converts the initial health score based on the original assessment data, it performs the following conversion process: First, the system subtracts the theoretical minimum extreme value of the corresponding dimension from the original scores of the five dimensions included in the multi-dimensional quantitative evaluation task, and then performs a division ratio operation with the range of the corresponding dimension (i.e., the difference between the full score extreme value and the theoretical minimum extreme value). The calculation result is then mapped to the percentage benchmark range to obtain the standard score of each dimension.

[0027] In the physical health dimension, the assessment item array contains 36 assessment items, corresponding to four secondary indicator variables: physical pain, energy, physiological function, and physical role limitations. Each item is quantified and assigned a discrete integer value from 1 to 5. The module calculates the sum of the values ​​of the 36 items to obtain the raw score for physical health, with the maximum score for this dimension set at 180 points.

[0028] In the daily living skills dimension, the assessment item array includes six core items: eating, washing, dressing, toileting, walking, and climbing stairs. Each item is quantified and assigned a value on a scale of 0, 5, 10, and 15. The module calculates the sum of the values ​​for these six items to obtain the raw score for daily living skills, with a maximum score of 100 points for this dimension.

[0029] In the psychological state dimension, the assessment item array includes two subsets: anxiety and depression, each containing 20 items. Each item is quantified and assigned a discrete integer value from 1 to 4. The module summarizes the scores of all items to obtain the raw score, and then calculates the psychological state standard score using a preset linear mapping operator (standard score = raw score × 1.25), with the maximum score set at 100.

[0030] In the cognitive function dimension, the assessment item array includes five subsets: orientation, memory, attention, calculation, and language ability, totaling 30 assessment items. Each item uses a binary assignment method, assigning a value of 1 if the condition is met and 0 if the condition is not met. The module calculates the sum of the values ​​of these 30 items to obtain the raw score for cognitive function, with a maximum score of 30.

[0031] In the social functioning dimension, the assessment item array includes three secondary indicators: social interaction, work or learning ability, and family role adaptation, with a total of 18 assessment items. Each item is quantified and assigned a discrete integer value from 1 to 5. The module calculates the raw score for social functioning, with a maximum score of 90.

[0032] After obtaining the raw scores for each dimension, the module calculates the standardized scores for each dimension using the following first mathematical model: ; in, Indicates the first Standard scores for each dimension Indicates the first The original scores of each dimension, Indicates the first The maximum score in each dimension Indicates the first The theoretical minimum extreme value of each dimension, These correspond to the five dimensions mentioned above.

[0033] Secondly, the system obtains the weight coefficients independently configured for each of the aforementioned dimensions, multiplies the standard score of each dimension by its corresponding weight coefficient, and sums all the product results to obtain the preliminary health score. The weight coefficients for each dimension are given by a preset configuration matrix, and the weight coefficients are set accordingly. satisfy And the constraint is .

[0034] Preliminary health score The calculation uses the following second mathematical model: ; in, A preliminary health score; For the aforementioned Standard scores for each dimension For the aforementioned The weighting coefficients for each dimension. Further, the answering behavior indicators include single-question answering time, overall missed question rate, and logical consistency of answers; when the five-dimensional health assessment module uses the answering behavior indicators to generate correction coefficients for dynamic multiplier calibration, it specifically executes the following logic: The five-dimensional health assessment module has an embedded behavioral indicator collection thread. As the target user answers each assessment item, the system background calls the system clock to record the page loading completion timestamp. Timestamp of receiving confirmation of submission Calculate the time taken to answer a single question. .in, The time spent answering a single question. The timestamp of receiving the confirmation submission instruction. This is the timestamp indicating when the page finished loading. Simultaneously, the system counts the number of unassigned entries in the currently submitted question group, divides this number by the total number of current response entries, and calculates the overall missed question rate. Furthermore, the system pre-sets logically related verification question pairs in the evaluation item array (e.g., the meaning of question A is positively or negatively correlated with the meaning of question B). The system calculates the Pearson correlation coefficient between the scores of the verification question pairs to measure the logical consistency of the answers. .

[0035] When it is detected that the time taken to answer a single question exceeds a preset time threshold. The overall missing question rate exceeds a preset proportion. Or the logical consistency of the answer is lower than the preset benchmark. If at least one of the abnormal conditions occurs, it is determined that the current answering process is distorted due to fluctuations in the target user's cognitive state.

[0036] In this embodiment, the aforementioned threshold parameter is not arbitrarily set, but rather dynamically matched or fixed based on prior testing benchmarks of the elderly population in the community. Specifically, regarding the preset time consumption threshold... The system takes the average response time for that specific assessment item in the prior database of normal elderly people plus twice the standard deviation as the benchmark (in practice, it is usually within the range of 15 to 30 seconds); the preset ratio The preset benchmark is set to a fixed value of 15%. To verify the lower limit of the absolute value of the Pearson correlation coefficient between the question and the answer choices, it is set to 0.6. When the absolute value of the calculated correlation coefficient is lower than 0.6, it is determined that there is a logical contradiction in the answer.

[0037] The preset penalty attenuation parameter is retrieved as the correction coefficient based on the corresponding triggered abnormal situation. The mapping table maintained internally by the system contains specific tiered penalty logic: if only one of the above three abnormal conditions is triggered (e.g., only excessive time consumption, but no questions missed and the logic is consistent), then the correction coefficient... The value is assigned to 0.1; if any two abnormal conditions are triggered simultaneously, then the correction coefficient is... The value is assigned to 0.15; if all three abnormal conditions are triggered simultaneously, indicating that the user's current state is extremely poor, then the correction coefficient is... The value is assigned to the extreme value of 0.2. This precise step-like mapping effectively avoids excessive dimensionality reduction and score deduction due to single operational errors by elderly users.

[0038] The system multiplies the correction coefficient with the preliminary health score to output the target health score after dimensionality reduction adjustment. The calculation process uses the following third mathematical model: ; in, To score for target health, Based on the aforementioned preliminary health score, This refers to the aforementioned correction factor.

[0039] Specifically, the breakpoint recovery and data synchronization module is communicatively connected to the five-dimensional health assessment module. It is used to save the original assessment data in the form of fragments to the local cache in real time and perform incremental cloud backup. When an abnormal assessment interruption is detected and the system is reconnected, the latest data status is locked through a two-way timestamp verification mechanism between the local and cloud systems, and the breakpoint recovery operation is triggered.

[0040] Furthermore, when the breakpoint recovery and data synchronization module locks the latest data status through a two-way timestamp verification mechanism and triggers the breakpoint recovery operation, it executes the following control logic: The system breaks down the multi-dimensional quantitative assessment task into independent assessment units. Each time the target user completes an assessment item, the system immediately updates the fragmented data and progress status identifier in the local cache and performs incremental cloud backup using an asynchronous thread. Specifically, the system generates a JSON data packet for each assessment item, containing a question number identifier, option key value, single question score, and current timestamp. Upon receiving a touch command, this data packet is synchronously written to the device's local SQLite database or LocalStorage, and a background asynchronous transmission queue, independent of the main user interface thread, is activated. The JSON data packet and the index of the currently completed maximum question number are then pushed to the cloud database for incremental overwriting.

[0041] Once the abnormal interruption (such as application process being terminated by the system or network physical link being disconnected) is recovered, the initialization program starts and extracts the first latest modification timestamp from the local cache. With the second most recent modification timestamp in the cloud database .

[0042] The first latest modified timestamp and the second latest modified timestamp are compared numerically. The end with the closest time to the current time (i.e., the end with the largest timestamp value) is used as the reference data end. The data status (including the set of answered question data packets and the maximum question number index) in the reference data end is overwritten to the other end to ensure data consistency across multiple ends. A routing jump instruction is sent to the age-friendly interactive interface module to guide the age-friendly interactive interface module to jump to the first unfinished evaluation item corresponding to the maximum question number index plus one for rendering.

[0043] Specifically, the intelligent data analysis module is communicatively connected to the five-dimensional health assessment module and the breakpoint continuation assessment and data synchronization module, respectively, and is used to construct a time series fitting model based on the target user's target health score in previous assessments, and generate a risk warning signal with the corresponding danger level when it is determined that the target health score or its time series change characteristics meet the multi-level warning triggering conditions.

[0044] Furthermore, when the intelligent data analysis module constructs a time series fitting model based on the target user's target health scores from previous assessments, it specifically performs the following calculation steps: First, extract the target user's assessment data sets at different time points from the health records. .in, This is a collection of data from previous evaluations. Represents the first element in the set. Data points, For the first Each time point For the first The target health score or single-dimensional standard score vector at each time point. and It is a positive integer. The total number of valid evaluations for the target user, and The value is a positive integer greater than or equal to 2. The system filters and removes extreme values ​​caused by operational anomalies by setting a standard deviation threshold, thus obtaining a cleaned and effective dataset.

[0045] Secondly, regarding the difference between adjacent time points For sampling periods longer than the preset period, a linear interpolation algorithm is used to smoothly fill in the missing data at time points due to evaluation interruptions or omissions. The interpolation formula is: ; in, This represents the smoothed-fill value at the missing time point, calculated using linear interpolation. Indicates the missing time points where data needs to be filled in; For the first Time nodes The next adjacent valid time node, For the first Time nodes The next adjacent valid time node The corresponding target health score or single-dimensional standard score vector.

[0046] This generates a continuous healthy time series with uniform time intervals. .in, It is a continuous healthy time series. This represents the total number of consecutive time points generated after smoothing and numerical imputation. Represents consecutive time points distributed at equal time intervals. To be at the corresponding time node ( and The target health score or single-dimensional standard score vector on (where is a positive integer).

[0047] Finally, a linear regression algorithm is used to perform regression fitting on the health time series, generating trend feature data reflecting the health evolution trajectory in each dimension. The system constructs a linear regression model with time span as the independent variable and health score fitting value as the dependent variable. Calculate the slope parameter using the least squares method and intercept parameter .in, is the dependent variable in this linear regression model, representing the continuous fit score of the health evolution trajectory; In a linear regression model, represents the independent variable, indicating the time span variable; The slope The intercept. The slope. The numerical values ​​and their positive or negative values ​​are extracted as trend feature data to characterize the rate and direction of change of health indicators.

[0048] Furthermore, the multi-level early warning triggering conditions include static absolute value conditions, sequential cumulative decrease conditions, and single sudden drop conditions; the risk early warning signals are divided into first-level early warning, second-level early warning, and third-level early warning. The intelligent data analysis module is configured to execute the following condition judgment network when generating risk early warning signals: Condition 1 Judgment Logic: When the target health score or the score of any dimension in a single assessment... Falling into the pre-defined static low score interval At that time, based on the static absolute value condition, the risk warning signal corresponding to the hazard level is triggered, wherein... and These represent the lower and upper extreme values ​​between the static low-level partitions, respectively. The system maintains a mapping matrix, such as the target health score. Mapped to Level 1 warning Mapped to Level 2 warning Mapped to Level 3 warning.

[0049] Condition 2 Judgment Logic: When the score of any dimension shows a continuous decline in consecutive evaluation batches (i.e., based on the latest evaluation index) Based on the baseline, the scores at three adjacent time points satisfy ), and the cumulative decline The threshold values ​​are successively increased by three extreme value drops (such as a threshold set). ,satisfy When, based on the cumulative descent condition of the sequence, compare respectively The threshold set triggers the corresponding first-level warning, second-level warning, or third-level warning. The most recent occurrence in the time series (i.e., the th) (Evaluate the score at the corresponding time point) (time); For the previous time (the first) (Evaluate the score at the corresponding time point) (time); For the first two (the first two) (Evaluate the score at the corresponding time point) (time); This represents the cumulative decrease. This refers to the specific extreme value drop threshold parameter.

[0050] To ensure the objectivity and uniformity of early warning triggering, the set of extreme value drop thresholds... It is set as a fixed percentage value linked to the maximum score extreme value of the evaluation dimension. In practice, Set to 10% of the maximum score for the corresponding dimension. Set at 15% of the maximum score. Set to 20% of the maximum score.

[0051] Condition 3 Judgment Logic: When determining the difference in the single decrease of the score in any dimension between two adjacent assessments. Exceeding the preset threshold for sudden drop When this occurs, the system executes interrupt control, disabling the output judgment of the static absolute value condition, and directly triggering the risk warning signal of no less than the second-level warning based on the single sudden drop condition (i.e., the output result set is limited to the second or third level). Correspondingly, the sudden drop threshold... Set to 25% of the maximum score for the corresponding dimension.

[0052] Specifically, the personalized recommendation engine module is communicatively connected to the intelligent data analysis module and the user management module, respectively, and is used to obtain the target health score, the risk warning signal, the health record and historical suggestion feedback records. It uses a multi-dimensional weighted collaborative filtering matching algorithm to calculate the comprehensive matching degree of each candidate suggestion in order to generate personalized intervention suggestions for the target user.

[0053] Furthermore, when calculating the comprehensive matching degree of each candidate suggestion using a multi-dimensional weighted collaborative filtering matching algorithm, the personalized recommendation engine module executes the following vector calculation model: The system has a pre-set suggestion library, where each candidate suggestion is abstracted as a multi-dimensional feature vector. For each candidate suggestion in the pre-set suggestion library, cosine similarity or vector dot product is calculated based on its health dimension label feature value and the weakest dimension (i.e., the dimension whose score is lower than a pre-set benchmark) in the target health score to obtain the evaluation result matching degree. .

[0054] The user profile matching degree is calculated based on the applicable population tags of the candidate suggestions and the past medical history and age characteristics in the health records. The system encodes user profile data into user profile vectors and applicable audience tags into constraint vectors. It calculates the Hamming distance between the two and converts it into a numerical value using an inverse proportional function or feature mapping algorithm. The feature similarity score of the interval is used to unify the calculation direction and dimension of the matching degree; if there is a hard medical contraindication conflict, the matching degree of that item is set to an absolute zero value.

[0055] Based on the target user's valid markers (e.g., click-through rate, execution completion rate) for past intervention suggestions in the historical suggestion feedback records, a score is calculated using collaborative filtering or latent semantic modeling to obtain the historical feedback matching degree. .

[0056] Independent feature weight multipliers are assigned to the matching degree of the evaluation result, the matching degree of the user profile, and the matching degree of the historical feedback. (satisfy The three factors are multiplied by their corresponding feature weight multipliers and then linearly summed to obtain the overall matching degree of each candidate suggestion. The calculation formula is as follows: ; in, For overall matching accuracy, To determine the degree of matching of the aforementioned assessment results, For user profile matching degree, For historical feedback matching degree, , and This refers to the feature weight multiplier. In specific implementation scenarios, considering that the current health assessment score has the highest intervention priority, the system defaults to setting the feature weight multiplier to a fixed empirical value: , , Furthermore, when the target user is a cold-start user entering the system for the first time, i.e., lacking historical suggestion feedback records, the system will... Forced to zero, and for and Perform proportional normalization amplification (i.e., reassign values) This ensures that the sum of the weights is always 1.

[0057] Furthermore, the personalized recommendation engine module is also configured with a sequence reconstruction executor, used for: All candidate suggestions with the calculated comprehensive matching degree are sorted in descending order of numerical value to generate an initial recommendation sequence.

[0058] In response to the risk warning signal triggered by the intelligent data analysis module, the system extracts the abnormal dimension identifier associated with the risk warning signal. The system iterates through the initial recommendation sequence to find candidate suggestions whose health dimension labels match the abnormal dimension identifier, and performs forced placement and rank promotion operations on these suggestions in the sorted sequence (i.e., forcibly modifying their index value to the first interval value of the sequence). After rank reconstruction, the system outputs the personalized intervention suggestion list data structure.

[0059] Specifically, the age-friendly interactive interface module is communicatively connected to the personalized recommendation engine module and the five-dimensional health assessment module, and is used to output the personalized intervention suggestions and the risk warning signals in a visual rendering format that exceeds the preset benchmark parameters.

[0060] Furthermore, when the age-friendly interactive interface module outputs the risk warning signal in a visual rendering format that exceeds preset baseline parameters, it is configured as follows: The front-end DOM node attributes are rendered using a larger font structure than the conventional baseline threshold (e.g., 1.5 times the default font size of the operating system or a font size setting of no less than 18pt) and a high-contrast color scheme (e.g., the contrast ratio of the background color to the foreground color calculated by RGB is no less than 4.5:1).

[0061] In response to the received risk warning signals of different hazard levels, different levels of alert coloring channels are invoked in the visualization chart components (such as radar charts or line charts rendered based on Canvas or SVG) to dynamically highlight the corresponding coordinate points or areas of the abnormal dimension. For example, the third-level warning uses the RGB(255,0,0) color channel with a periodic transparency gradient animation, the second-level warning uses the RGB(255,165,0) color channel, and the first-level warning uses the RGB(255,255,0) color channel. Simultaneously, the underlying media control interface of the operating system is invoked to input the text data of personalized intervention suggestions into the TTS (Text-To-Speech) speech synthesis engine, and a progressive intervention guidance voice broadcast stream is launched synchronously.

[0062] Based on the above system architecture, embodiments of the present invention also provide a method for intelligent health assessment and management of elderly people in communities, such as... Figure 2 As shown, the method includes the following steps: Step S1: Establish and encrypt the target user's health record. The system allocates memory space to load the authentication module and receives the target user's identification data stream. After parsing the data stream, a corresponding row-level storage structure is allocated in the server database cluster. The target user's attribute parameters (including basic personal information, past medical history, medication details, physical examination records, etc.) are obtained through the input / output interface. Subsequently, the system's main processor calls symmetric encryption algorithm instruction sets such as AES, uses an independently distributed key to convert the above attribute parameters into a ciphertext binary stream, and writes the ciphertext binary stream into a designated sector of the physical storage medium, completing the establishment and encrypted persistence of the health record.

[0063] Step S2: Provide the target user with a multi-dimensional quantitative assessment task to obtain raw assessment data. During the execution of the assessment task, collect the target user's answer behavior indicators in real time. Use the answer behavior indicators to generate correction coefficients to dynamically multiply and calibrate the preliminary health score calculated based on the raw assessment data, and output the target health score. The system sequentially extracts metadata from the database for five categories of question banks: physical health, daily living skills, psychological state, cognitive function, and social function. Encapsulates this metadata into a JSON message and sends it to the client for rendering. While monitoring user touch events to collect raw assessment data (values ​​of each question option), the client uses the system clock register to record the absolute timestamp of each event trigger. Single-question answering time data is generated by calculating the difference between adjacent timestamps, overall missing question rate data is generated by statistically analyzing unassigned variables, and answer logical consistency data is generated by executing a cross-comparison logic script. These behavioral indicator data are sent back to the server, which compares them with the data in the threshold register according to a preset comparison instruction. If the abnormal conditions are met, the server queries the penalty coefficient mapping table to obtain the correction coefficient variable, performs a floating-point multiplication operation with the preliminary health score variable calculated according to the weight, and calculates the target health score after dynamic calibration.

[0064] Step S3: The original evaluation data is saved to the local cache in real time in fragment form and incrementally backed up to the cloud. This allows for locking the latest data state and triggering breakpoint recovery operations when an evaluation interruption is detected and the system reconnects, via a two-way timestamp verification mechanism between the local and cloud servers. Each time a valid option key value is input, the system immediately performs a write operation to store the current question number index and option value in local non-volatile memory (such as Flash Memory) and encapsulates it into a data packet containing a UNIX timestamp. The network controller sends this data packet to the cloud server queue via the TCP / IP protocol stack in an asynchronous, non-blocking manner. When the system main control program detects a restart or session recovery event, the system reads the timestamp of the data packet stored in the local Flash and sends a request to obtain the timestamp of the last confirmed record in the cloud. The main processor uses the arithmetic logic unit to compare the absolute values ​​of the two timestamps and uses the record corresponding to the timestamp with the larger value as the final state benchmark. If the cloud timestamp is larger, the cloud data packet is downloaded via the network interface and overwrites the local storage area; conversely, the smaller timestamp is used. After synchronization is complete, the memory instruction pointer is pointed to the address after the largest question number index, and the rendering task of the evaluation page continues.

[0065] Step S4: Construct a time series fitting model based on the target user's target health scores from previous assessments, and generate a risk warning signal corresponding to the corresponding danger level when the target health score or its time series change characteristics meet the multi-level warning trigger conditions. The system executes a SELECT operation in the database to retrieve the target health score array corresponding to a specific user ID in ascending order of the time axis. The central processing unit executes a linear interpolation subroutine to complete the array elements of discontinuous values ​​on the time axis. Then, the linear regression algorithm library function is executed to iteratively calculate the mean square error and obtain the slope vector and intercept vector of the fitted line. The system comparison unit determines whether the current array element or slope vector meets the trigger conditions based on the multi-condition branch structure. If the score is lower than the constant value in the static threshold memory area, or the score decreases continuously and the cumulative difference is greater than the cumulative drop constant value, or the difference between two adjacent scores is greater than the single drop constant value, the system sets the risk status register and broadcasts a risk warning interrupt signal containing the corresponding warning level (first, second, or third level) to the system bus.

[0066] Step S5: Obtain the target health score, the risk warning signal, the health record, and historical suggestion feedback records. Calculate the comprehensive matching degree of each candidate suggestion using a multi-dimensional weighted collaborative filtering matching algorithm to generate personalized intervention suggestions for the target user. The system extracts the above information to construct a user feature matrix set and simultaneously loads the system-wide candidate suggestion feature matrix set. The arithmetic unit performs matrix multiplication and dot product operations to calculate a floating-point scalar matrix representing the matching degree of the evaluation result, a floating-point scalar matrix representing the matching degree of the user record, and a floating-point scalar matrix representing the matching degree of historical feedback, respectively. Perform scalar multiplication of the above three matrices with their corresponding weight scalar constants, followed by matrix addition, to obtain the comprehensive matching degree. The sorting algorithm unit (implemented as quicksort or mergesort) rearranges the matrix in descending order according to the values ​​of the comprehensive matching degree column. Subsequently, the judgment logic unit detects the presence of a high-level risk warning signal flag. If present, it traverses the rearranged matrix and moves the row data pointer matching the abnormal dimension directly to the beginning position of the matrix storage area. Finally, the set of candidate suggestion data blocks after pointer rearrangement is output.

[0067] Step S6: Output the personalized intervention suggestion and the risk warning signal in a visual rendering format that exceeds the preset baseline parameters. The graphics processing unit of the terminal device receives the data block set output in step S5 and renders it using CSS or style objects of the corresponding UI framework. During this process, the font size attribute variable is overwritten with a value greater than the preset lower limit, and the foreground and background color HEX values ​​are overwritten with values ​​that satisfy the high contrast test formula. When a data block containing the risk warning signal flag is parsed, a rendering instruction containing specific RGB color values ​​and frame animation parameters are sent to the graphics pipeline to drive the screen pixel array to blink or highlight in the corresponding chart coordinate area. In addition, the audio processing module receives the suggestion text string, converts it into a PCM digital audio stream using the internal audio decoding library, and drives the speaker hardware to play voice prompts after passing through a digital-to-analog converter (DAC).

[0068] To more clearly illustrate the internal data flow process of the above systems and methods, the following provides an analysis of the execution process in several specific implementation scenarios: In Scenario 1, the system establishes and verifies a connection session with target user A (User_ID: 10001). User A's profile data vector contains the tag "essential hypertension". During the multi-dimensional quantitative assessment in step S2, the system issues an assessment instruction set for the cognitive function dimension. During the question-answering process, the system's behavior monitoring clock records the time user A spends on question 24. It is much greater than the preset time threshold. Furthermore, the logical consistency is demonstrated by taking the inner product of the option vectors given in the subsequent two logic verification questions. Below the preset benchmark The calculation unit generates a preliminary health score. After the division, since the anomaly detection logic is met, the system retrieves the penalty decay parameter. The arithmetic unit performs floating-point multiplication. The system updated the target health score to 61.2.

[0069] In Scenario 2, for user B (User_ID: 10002), during step S4, the time series extraction module retrieves the standard score vectors of user B's psychological state for four consecutive months. The comparison unit, during the conditional judgment process, detected the difference in single decrease values ​​between two adjacent transactions. The sudden drop exceeded the preset threshold. At this point, the logic controller triggers a single sudden drop condition, disables the conventional static interval judgment logic, directly generates a risk warning signal data frame with the corresponding level of "Level 2 Warning", and writes it to the system cache.

[0070] In Scenario 3, for the comprehensive matching calculation of user A mentioned above, the recommendation engine loads candidate suggestions. (Tags include: high-intensity anaerobic exercise). In the matrix operation phase of step S5, although the evaluation result matching degree calculation was assigned a high initial similarity scalar due to its low physical health score, when calculating the user profile matching degree, the system feature comparison engine detected the string "essential hypertension" in the user profile, which collided with the contraindication tag included in the suggestion. The system immediately rejected the candidate suggestion. The user profile matching score component is forcibly assigned a value of 0.0. This will be applied in subsequent overlays. In the linear operation of weights, the overall matching degree of this suggestion The record pointer was significantly lowered, and after the quicksort algorithm was executed, it was moved to the end of the sorted sequence, thus avoiding the output of unsafe health intervention suggestions.

[0071] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0072] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

Claims

1. A smart health assessment and management system for elderly people in a community, characterized in that, The system includes: The user management module is used to create health records for target users and store them in encrypted form. The five-dimensional health assessment module is connected to the user management module and is used to provide the target user with a multi-dimensional quantitative assessment task to obtain raw assessment data. During the execution of the assessment task, the module collects the target user's answer behavior indicators in real time, uses the answer behavior indicators to generate correction coefficients, and performs dynamic multiplier calibration on the preliminary health score calculated based on the raw assessment data to output the target health score. The breakpoint recovery and data synchronization module is connected to the five-dimensional health assessment module. It is used to save the original assessment data to the local cache in real time in the form of fragments and perform incremental cloud backup. When an abnormal assessment interruption is detected and the connection is restored, the latest data status is locked through the two-way timestamp verification mechanism between the local and cloud and the breakpoint recovery operation is triggered. The intelligent data analysis module is connected to the five-dimensional health assessment module and the breakpoint continuation assessment and data synchronization module, respectively. It is used to construct a time series fitting model based on the target health score of the target user in previous assessments, and generate a risk warning signal with the corresponding danger level when it is determined that the target health score or its time series change characteristics meet the multi-level warning triggering conditions. The personalized recommendation engine module is communicatively connected to the intelligent data analysis module and the user management module, respectively, and is used to obtain the target health score, the risk warning signal, the health record and historical suggestion feedback records. It uses a multi-dimensional weighted collaborative filtering matching algorithm to calculate the comprehensive matching degree of each candidate suggestion in order to generate personalized intervention suggestions for the target user. The age-friendly interactive interface module is communicatively connected to the personalized recommendation engine module and the five-dimensional health assessment module, respectively, and is used to output the personalized intervention suggestions and the risk warning signals in a visual rendering format that exceeds the preset benchmark parameters. The answering behavior indicators include time spent answering a single question, overall missed question rate, and logical consistency of answers; the five-dimensional health assessment module, when using the answering behavior indicators to generate correction coefficients for dynamic multiplier calibration, is specifically used for: When at least one of the following abnormal situations is detected: the time taken to answer a single question is higher than a preset time threshold, the overall missing question rate exceeds a preset proportion, or the logical consistency of the answer is lower than a preset benchmark, it is determined that the current answering process is distorted due to fluctuations in the cognitive state of the target user. According to the corresponding triggered abnormal situation, the preset penalty decay parameter is retrieved as the correction coefficient, and the correction coefficient is multiplied and converted with the preliminary health score to output the target health score after dimensionality reduction adjustment. Specifically, when the breakpoint recovery and data synchronization module locks the latest data status and triggers the breakpoint recovery operation through a two-way timestamp verification mechanism, it is used for: The multi-dimensional quantitative evaluation task is divided into independent evaluation units. When the target user completes each evaluation item, the fragment data and progress status identifier in the local cache are updated immediately, and the incremental cloud backup is executed using an asynchronous thread. After the abnormal interruption is recovered, the first latest modified timestamp in the local cache and the second latest modified timestamp in the cloud database are extracted respectively. The first latest modified timestamp is compared with the second latest modified timestamp. The one with the closest time to the present is used as the reference data end. The data status in the reference data end is overwritten to the other end, and the age-friendly interactive interface module is guided to jump to the first incomplete evaluation item.

2. The intelligent health assessment and management system for the elderly in the community according to claim 1, characterized in that, The five-dimensional health assessment module, when converting the initial health score based on the original assessment data, is specifically used for: The raw scores of the five dimensions of physical health, daily living ability, mental state, cognitive function and social function included in the multi-dimensional quantitative assessment task are subtracted from the theoretical minimum extreme value of the corresponding dimension, and then divided by the range of the corresponding dimension. The calculation results are then mapped to the percentage benchmark range to obtain the standard score of each dimension. Obtain the weight coefficients that are independently configured for each of the above dimensions, multiply the standard score of each dimension by its corresponding weight coefficient, and sum all the product results to obtain the preliminary health score.

3. The intelligent health assessment and management system for the elderly in the community according to claim 1, characterized in that, When the intelligent data analysis module constructs a time series fitting model based on the target user's target health scores from previous assessments, it is specifically used for: Extract the target user's assessment data at different time points from the health record, and filter and remove extreme scores caused by abnormal operation. Linear interpolation algorithms are used to smoothly fill in missing time point data caused by evaluation interruption or omission, so as to generate a continuous health time series. The health time series is fitted using a linear regression algorithm to generate trend feature data that reflects the health evolution trajectory in each dimension.

4. The intelligent health assessment and management system for the elderly in the community according to claim 1, characterized in that, The multi-level early warning triggering conditions include static absolute value conditions, sequential cumulative decline conditions, and single sudden drop conditions; the risk early warning signals are divided into first-level early warning, second-level early warning, and third-level early warning. The intelligent data analysis module is configured to generate risk warning signals as follows: When the target health score or the score of any dimension in a single assessment falls into a predefined static low score range, the risk warning signal corresponding to the danger level is triggered based on the static absolute value condition. When the score of any dimension shows a continuous decline in consecutive evaluation batches and the cumulative decline reaches three increasing extreme value thresholds in sequence, the first level warning, the second level warning, or the third level warning is triggered based on the cumulative decline condition of the sequence. When it is determined that the difference between the single decrease in the score of any dimension in two adjacent assessments exceeds the preset sudden drop threshold, the static absolute value condition is masked, and the risk warning signal of no less than the second level warning is directly triggered based on the single sudden drop condition.

5. The intelligent health assessment and management system for the elderly in the community according to claim 1, characterized in that, The personalized recommendation engine module, when calculating the comprehensive matching degree of each candidate suggestion using a multi-dimensional weighted collaborative filtering matching algorithm, is specifically used for: For each candidate suggestion in the preset suggestion library, the matching degree of the evaluation result is calculated based on its health dimension label and the weakness dimension in the target health score; The user profile matching degree is calculated based on the applicable population tags of the candidate suggestions and the past medical history and age characteristics in the health records. The historical feedback matching degree is calculated based on the target user's effective marking of past intervention suggestions in the historical suggestion feedback record. Independent feature weight multipliers are assigned to the matching degree of the evaluation result, the matching degree of the user profile, and the matching degree of the historical feedback. The three are multiplied by their corresponding feature weight multipliers and then linearly superimposed to obtain the comprehensive matching degree of each candidate suggestion.

6. The intelligent health assessment and management system for the elderly in the community according to claim 5, characterized in that, The personalized recommendation engine module is also used for: All candidate suggestions with the calculated comprehensive matching degree are sorted in descending order; In response to the risk warning signal triggered by the intelligent data analysis module, the candidate suggestions that match the abnormal dimension that triggered the warning are forced to the top and their rank is promoted in the sorting sequence. After rank reconstruction, the personalized intervention suggestions are output.

7. The intelligent health assessment and management system for the elderly in the community according to claim 1, characterized in that, When the age-friendly interactive interface module outputs the risk warning signal in a visual rendering format that exceeds preset baseline parameters, it is configured as follows: The interface rendering employs a larger font structure than the conventional baseline threshold and a high-contrast color scheme. In response to the risk warning signals of different hazard levels, different levels of alert coloring channels are invoked in the visualization chart to highlight and dynamically annotate the abnormal dimensions, and a progressive intervention guidance voice broadcast stream is launched simultaneously.

8. A method for intelligent health assessment and management of elderly people in a community, characterized in that, The method includes the following steps: Step S1: Establish and encrypt the health records of the target users; Step S2: Provide the target user with a multi-dimensional quantitative assessment task to obtain raw assessment data, and collect the target user's answer behavior indicators in real time during the execution of the assessment task. Use the answer behavior indicators to generate correction coefficients to dynamically multiply and calibrate the preliminary health score calculated based on the raw assessment data, and output the target health score. Step S3: Save the original evaluation data to the local cache in real time in the form of fragments and perform incremental cloud backup, so that when an evaluation abnormal interruption is detected and reconnected, the latest data status is locked through the two-way timestamp verification mechanism between the local and cloud and the breakpoint recovery operation is triggered. Step S4: Construct a time series fitting model based on the target user's target health score in previous assessments, and generate a risk warning signal of the corresponding danger level when the target health score or its time series change characteristics meet the multi-level warning triggering conditions; Step S5: Obtain the target health score, the risk warning signal, the health record, and the historical suggestion feedback record; use a multi-dimensional weighted collaborative filtering matching algorithm to calculate the comprehensive matching degree of each candidate suggestion to generate personalized intervention suggestions for the target user. Step S6: Output the personalized intervention suggestion and the risk warning signal in a visual rendering format that exceeds the preset baseline parameters; The answering behavior indicators include time spent answering a single question, overall missed question rate, and logical consistency of answers; when the five-dimensional health assessment module uses the answering behavior indicators to generate correction coefficients for dynamic multiplier calibration, it is specifically used for: When at least one of the following abnormal situations is detected: the time taken to answer a single question is higher than a preset time threshold, the overall missing question rate exceeds a preset proportion, or the logical consistency of the answer is lower than a preset benchmark, it is determined that the current answering process is distorted due to fluctuations in the cognitive state of the target user. According to the corresponding triggered abnormal situation, the preset penalty decay parameter is retrieved as the correction coefficient, and the correction coefficient is multiplied and converted with the preliminary health score to output the target health score after dimensionality reduction adjustment. Specifically, when the breakpoint recovery and data synchronization module locks the latest data status and triggers the breakpoint recovery operation through a two-way timestamp verification mechanism, it is used for: The multi-dimensional quantitative evaluation task is divided into independent evaluation units. When the target user completes each evaluation item, the fragment data and progress status identifier in the local cache are updated immediately, and the incremental cloud backup is executed using an asynchronous thread. After the abnormal interruption is recovered, the first latest modified timestamp in the local cache and the second latest modified timestamp in the cloud database are extracted respectively. The first latest modified timestamp is compared with the second latest modified timestamp. The end with the closest time to the present is used as the reference data end. The data status in the reference data end is overwritten to the other end, and the age-friendly interactive interface module is guided to jump to the first incomplete evaluation item.

Images