Diabetes Management System Based on Big Data

Abstract

This invention relates to the field of health data management technology, specifically a diabetes management system based on big data. The system includes an anomaly monitoring module, a risk analysis module, a health status simulation module, and a monitoring and management module. This invention significantly improves diabetes management through data processing and analysis technologies. Refined data review and anomaly monitoring strategies ensure data quality, eliminating unqualified data, identifying anomalies, and analyzing the development trends of anomalies through time series analysis. This data processing not only improves data accuracy but also enhances the predictive ability of potential health risks. The assessment and classification of risk areas allow healthcare providers to more accurately understand and prevent disease risks. By simulating the health status of diabetic patients, future health changes can be predicted, helping to develop more proactive recovery plans, improving the efficiency and effectiveness of disease management, and promoting patients' quality of life.

Description

Technical Field

[0001] This invention relates to the field of health data management technology, and in particular to a diabetes management system based on big data. Background Technology

[0002] The field of health data management technology focuses on the effective collection, storage, management, and analysis of health-related data. This includes various technologies ranging from the implementation of electronic health record (EHR) systems to big data analytics, aiming to improve the quality of healthcare services, reduce costs, promote personalized medicine, and enhance patient management efficiency. By leveraging advanced data processing technologies such as machine learning and artificial intelligence, health data management can reveal statistical trends in disease causes and enable real-time monitoring of patient health, playing a crucial role in the prevention of various diseases, especially chronic diseases.

[0003] A diabetes management system is a technological solution specifically designed to monitor and process the health information of diabetic patients. This system helps healthcare professionals achieve more precise control and management of diabetes by collecting information such as blood glucose levels, dietary habits, and physical activity. Its uses include tracking disease progression, predicting blood glucose changes, and providing patients with customized health advice, significantly improving the quality of life and recovery outcomes for diabetic patients.

[0004] Existing technologies have significant shortcomings in practical application, particularly in data management and risk prediction. Data processing in existing systems is not deep enough, and the identification and analysis of anomalies are not precise enough, resulting in limited data quality and usability. Existing technologies also have limited capabilities in predicting health status and monitoring disease progression, failing to provide sufficient forward-looking and recovery recommendations. These deficiencies lead to slow disease management responses, hindering effective prevention of disease risks or adjustments to recovery strategies to address changes in patient health. For example, in long-term diabetes management, insufficient early warning of blood glucose spikes or complications leads to worsening of the patient's condition, increasing recovery costs and complexity. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a diabetes management system based on big data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A diabetes management system based on big data includes:

[0007] The anomaly monitoring module collects blood glucose, blood pressure, and weight data from diabetic patients, verifies the completeness of the data, removes unqualified data, identifies and analyzes anomalies in the data to obtain anomaly identification results, and performs time series analysis on the anomaly identification results to analyze the development trend of the abnormal data and obtain time trend analysis results.

[0008] Based on the time trend analysis results, the risk analysis module analyzes the clustering characteristics of abnormal data, identifies the clustering distribution, generates an initial risk region, performs a risk assessment on the initial risk region, identifies the risk level of diabetes symptoms, and obtains risk assessment indicators.

[0009] The health status simulation module uses the aforementioned risk assessment indicators to dynamically simulate the health status of diabetic patients, analyzes the differentiated health status transition probabilities, constructs a health transition probability matrix, uses the health transition probability matrix to predict the patient's health status in the future time period, evaluates the consistency of the prediction, and obtains the health status prediction results.

[0010] Based on the health status prediction results, the monitoring and management module identifies key monitoring indicators and intervention time periods, optimizes health monitoring parameters, generates a draft health monitoring plan for diabetic patients, adjusts the monitoring frequency, evaluates the effectiveness and impact of implementing the draft health monitoring plan for diabetic patients, and obtains a health management plan for diabetic patients.

[0011] As a further aspect of the present invention, the step of obtaining the anomaly identification result specifically includes:

[0012] Collect blood glucose, blood pressure, and weight data from diabetic patients, verify the integrity of the data, filter data that meets quality standards, and generate a quality-filtered dataset.

[0013] Anomaly identification is performed on the dataset after quality screening. By calculating the weighted Euclidean distance between each data point and its neighboring data points and comparing it with a dynamically adjusted threshold, data points that deviate from the normal range are identified, and an anomaly data point set is generated.

[0014] Time series analysis was performed on the aforementioned abnormal data point set to analyze the data development trend, according to the formula:

[0015]

[0016] Calculate the time-weighted moving average of the outliers to obtain the anomaly identification results;

[0017] Where E(t) represents the abnormal trend at time point t, a i (t) represents the i-th outlier data point at time t, w i The time weight represents the i-th data point at time t, and n is the number of outlier data points.

[0018] As a further aspect of the present invention, the steps for obtaining the time trend analysis results are specifically as follows:

[0019] Using the anomaly identification results, extract the timestamp of the anomaly occurrence, record the time of the anomaly occurrence, track the time of the anomaly occurrence in real time, and obtain a list of timestamps for the anomaly values;

[0020] A time window analysis was performed on the time-stamped list of outliers. By statistically analyzing the daily and weekly data, the frequency of outliers within the differentiated time window was evaluated, the outlier frequency was calculated, and the time window outlier frequency result was obtained.

[0021] Based on the anomaly frequency results within the aforementioned time window, the development and changes of the anomaly are assessed using the following formula:

[0022]

[0023] Calculate the outlier trend index R within the differential time window. b The time trend analysis results are obtained;

[0024] Where b represents the time variable, β0, β1, and γ are regression coefficients, ∈ b For the random error term, α and δ are adjustment coefficients, and e is the natural constant.

[0025] As a further aspect of the present invention, the step of obtaining the initial risk region specifically includes:

[0026] Based on the time trend analysis results, time periods with data fluctuations exceeding the threshold are filtered out to identify potential risk events and obtain records of potential risk time periods.

[0027] Based on the records of the potential risk time periods, the time periods are grouped, and risk patterns of different types are identified to obtain a risk pattern group set;

[0028] Based on the risk pattern grouping set, the formula is used:

[0029]

[0030] Calculate the risk value R for each group. c If the risk value exceeds the threshold, it is marked to obtain the initial risk area;

[0031] Among them, w j It is the risk weight, x cj is the risk value of the j-th indicator in the c-th group, and m is the number of indicators.

[0032] As a further aspect of the present invention, the steps for obtaining the risk assessment indicators are specifically as follows:

[0033] Key data indicators of diabetic patients are extracted from the initial risk area, including blood glucose levels, blood pressure and weight records of diabetic patients, to obtain a set of key data indicators.

[0034] Statistical analysis is performed on the set of key data indicators to calculate the mean and standard deviation of the differentiated key data indicators, assess the risk contribution of each indicator, and obtain the statistical analysis results.

[0035] Using the statistical analysis results, the criticality of the indicators is measured based on the volatility of the data, using the following formula:

[0036]

[0037] Calculate the coefficient of variation V for each indicator. k Risk assessment indicators are obtained;

[0038] Where, x k is a single data point, μ and σ are the mean and standard deviation of the indicator, respectively, and p is the number of indicators k.

[0039] As a further aspect of the present invention, the step of obtaining the health transition probability matrix specifically includes:

[0040] Using the aforementioned risk assessment indicators, including blood glucose control, weight change, and records of diabetic complications, the risk assessment indicators are standardized and converted into comparable scoring scales to obtain standardized risk scores.

[0041] By combining the standardized risk score with the health data of diabetic patients, the correlation is analyzed using the following formula:

[0042]

[0043] Estimate the transition probability P between different health states or Construct a health transition probability matrix;

[0044] Among them, S or α represents the standardized risk score indicating a transition from health state o to health state r. o β u and β r It is the adjustment coefficient, exp is the natural exponential function, and S ou A standardized risk score representing a transition from health state o to health state u.

[0045] As a further aspect of the present invention, the step of obtaining the health status prediction result specifically includes:

[0046] Using the health transition probability matrix, the applicability and consistency of predicting the health status of diabetic patients in the future time period are evaluated through data fit analysis of the health transition probability matrix, and a fit score record is obtained.

[0047] Based on the aforementioned fit score records, combined with the real-time health data of diabetic patients, the following formula is used:

[0048]

[0049] Calculate the probability prediction value of health status to obtain the health status prediction result;

[0050] Where P(t+1) is the predicted health status probability at the next time point, P(t) is the health status probability at the real-time time point, and x j (t) is the j-th indicator at time t, β j is the regression coefficient, B is the smoothing parameter, and m is the number of indicators.

[0051] As a further aspect of the present invention, the steps for obtaining the health management plan for diabetic patients specifically include:

[0052] Using the health status prediction results, analyze the risk level and potential health problems of diabetic patients, identify key monitoring indicators and necessary intervention periods, and generate a draft health monitoring plan for diabetic patients.

[0053] Implement the aforementioned draft health monitoring plan for diabetic patients, adjust the monitoring frequency, and use the following formula:

[0054] R adj =R base ·(1+α·P risk );

[0055] To assess the effectiveness and impact of the draft health monitoring plan for diabetic patients and to obtain a health management plan for diabetic patients;

[0056] Among them, R adj R represents the adjusted monitoring frequency. base This is the basic monitoring frequency, α is the adjustment coefficient, and P... risk It is a risk score for patients with diabetes.

[0057] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0058] This invention significantly improves diabetes management through data processing and analysis technologies. Refined data review and anomaly monitoring strategies ensure data quality, eliminating substandard data, identifying outliers, and analyzing their development trends through time series analysis. This data processing not only improves data accuracy but also enhances the ability to predict potential health risks. The assessment and classification of risk areas allows healthcare providers to more accurately understand and prevent disease risks. By simulating the health status of diabetic patients, future health changes can be predicted, helping to develop more proactive recovery plans, improving the efficiency and effectiveness of disease management, and promoting a higher quality of life for patients. Attached Figure Description

[0059] Figure 1 This is a system flowchart of the present invention;

[0060] Figure 2 This is a flowchart of the anomaly identification results in this invention;

[0061] Figure 3 This is a flowchart of the time trend analysis results in this invention;

[0062] Figure 4 This is a flowchart of the initialization of the risk area in this invention;

[0063] Figure 5 This is a flowchart of the risk assessment indicators in this invention;

[0064] Figure 6 This is a flowchart of the health transition probability matrix in this invention;

[0065] Figure 7 This is a flowchart of the health status prediction results in this invention;

[0066] Figure 8 This is a flowchart of the health management program for diabetic patients in this invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0068] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0069] Please see Figure 1 A big data-based diabetes management system includes:

[0070] The anomaly monitoring module collects blood glucose, blood pressure and weight data of diabetic patients, checks the integrity of the data, removes unqualified data, identifies and analyzes anomalies in the data, obtains anomaly identification results, and performs time series analysis on the anomaly identification results to analyze the development trend of abnormal data and obtain time trend analysis results.

[0071] The risk analysis module analyzes the clustering characteristics of abnormal data based on the time trend analysis results, identifies the clustering distribution, generates initial risk areas, performs risk assessment on the initial risk areas, identifies the risk level of diabetes symptoms, and obtains risk assessment indicators.

[0072] The health status simulation module uses risk assessment indicators to dynamically simulate the health status of diabetic patients, analyzes the differentiated health status transition probabilities, constructs a health transition probability matrix, uses the health transition probability matrix to predict the health status of patients in the future time period, evaluates the consistency of the prediction, and obtains the health status prediction results.

[0073] Based on the health status prediction results, the monitoring and management module identifies key monitoring indicators and intervention time periods, optimizes health monitoring parameters, generates a draft health monitoring plan for diabetic patients, adjusts the monitoring frequency, evaluates the effectiveness and impact of the implementation of the draft health monitoring plan for diabetic patients, and obtains a health management plan for diabetic patients.

[0074] Anomaly identification results include outlier identification, data integrity checks, and removal of non-compliant data. Time trend analysis results include data development trends, time series characteristics, and outlier analysis. Initialized risk areas include risk clustering characteristics, cluster distribution, and risk area assessment. Risk assessment indicators include risk level, risk assessment results, and cluster analysis results. The health transition probability matrix includes health status transition probability, model parameters, and dynamic simulation of health status. Health status prediction results include health status prediction for future time periods, prediction model performance, and health status transition analysis. The draft health monitoring plan for diabetic patients includes monitoring indicators, intervention time periods, and optimization of health monitoring parameters. The health management plan for diabetic patients includes adjustments to monitoring frequency, evaluation of health management effectiveness, and impact of monitoring plan implementation.

[0075] Please see Figure 2 The specific steps for obtaining the anomaly detection results are as follows:

[0076] Collect blood glucose, blood pressure, and weight data from diabetic patients, verify the integrity of the data, filter data that meets quality standards, and generate a quality-filtered dataset.

[0077] In the data collection phase of clinical research, collecting blood glucose, blood pressure, and weight data from diabetic patients is a crucial first step. This process involves regular monitoring using various medical devices, including blood glucose meters, blood pressure monitors, and electronic scales. Data validation and screening are performed using specialized data validation algorithms. These algorithms detect the completeness and accuracy of the data, automatically flagging any abnormal or significantly out-of-range readings. For example, blood glucose levels should not be below 3.9 mmol / L or above 7.8 mmol / L in non-postprandial states, and normal blood pressure should be between 90 / 60 and 120 / 80. Data cleaning and processing techniques are also used to screen data that meets medical quality standards, including removing any erroneous data points due to equipment malfunction or operational errors, generating a quality-screened dataset.

[0078] Outlier identification is performed on the quality-screened dataset by calculating the weighted Euclidean distance between each data point and its neighboring data points and comparing it with a dynamically adjusted threshold to identify data points that deviate from the normal range and generate an outlier data point set.

[0079] Detailed anomaly detection on the dataset is crucial to the accuracy and reliability of subsequent analysis. Anomaly detection is primarily performed by calculating the weighted Euclidean distance between each data point and its neighbors. This involves not only direct comparison of data points but also the introduction of dynamic thresholds to better accommodate individual physiological differences. Specifically, a neighborhood is defined for each data point, and the weighted distances between the target data point and its neighbors are calculated across three dimensions: blood glucose, blood pressure, and weight. The weights are determined based on the medical importance of each dimension and its contribution to diabetes control. For example, blood glucose data has a higher weight than weight data because blood glucose directly impacts daily diabetes management. This approach effectively identifies anomalous data points that deviate from the normal physiological range, laying a solid foundation for subsequent data analysis and generating an anomaly data point set.

[0080] Perform time series analysis on outlier data points to analyze data trends, according to the formula:

[0081]

[0082] Calculate the time-weighted moving average of the outliers to obtain the anomaly identification results;

[0083] Where E(t) represents the abnormal trend at time point t, a i (t) represents the i-th outlier data point at time t, w i The time weight of the i-th data point at time t is represented by n, where n is the number of outlier data points.

[0084] The time weighting is set based on the position and importance of the data point throughout the monitoring period. For example, in intraday blood glucose fluctuation monitoring, postprandial readings are given higher weight, which is especially crucial for understanding the patient's blood glucose control status.

[0085] Through actual data monitoring, a specific time point was set, and a total of 5 data points were monitored. The abnormal data points were identified as a(t) = [1, 0, 1, 1, 0].

[0086] The weights are set according to the importance of the data, with w = [0.5, 0.2, 0.5, 0.7, 0.1].

[0087] The calculation process is as follows:

[0088] Calculate the total weights:

[0089]

[0090] Calculate the weighted sum of outliers:

[0091]

[0092] Calculate E(t) using the formula:

[0093]

[0094] The results show that the weighted average of the abnormal trends at time point t is 0.85. The closer this value is to 1, the higher the degree of abnormality at that time point. This reflects the degree of abnormality and distribution of the abnormal data points, which helps to further analyze the development trend of the data and provide data support for medical intervention.

[0095] Please see Figure 3 The specific steps for obtaining the time trend analysis results are as follows:

[0096] Using the anomaly identification results, extract the timestamps of the anomalies, record the time of the anomalies, track the time of the anomalies in real time, and obtain a list of timestamps for the anomalies.

[0097] Identifying anomalies at various time points involves recording the time of occurrence of the anomalies, accurate to the specific timestamp, such as the year, month, day, hour, minute, and second of the occurrence. This ensures that each anomaly can be accurately tracked. The timestamp list includes the specific occurrence time of each anomaly, providing necessary basic data for subsequent analysis. Through efficient data indexing technology, data at any time point can be quickly accessed, which is crucial for processing batch time series data. Each timestamp is also associated with a measure of anomaly intensity, reflecting the severity of the anomaly and providing a basis for subsequent steps of anomaly frequency and trend analysis, resulting in a timestamp list of anomaly values.

[0098] Time window analysis is performed on the time-stamped list of outliers. By statistically analyzing the data day by day and week by week, the frequency of outliers within the differentiated time window is evaluated, the outlier frequency is calculated, and the outlier frequency result of the time window is obtained.

[0099] A detailed time window analysis is performed on the time-stamped list of outliers. By selecting an appropriate window length (e.g., one day or one week), the number of outliers occurring within each window is counted. This statistical analysis is accomplished through advanced data aggregation techniques, allowing for the rapid calculation of outlier frequencies from a batch of time stamps and the assessment of the temporal distribution trends of outliers. This time window analysis also includes the seasonality or periodicity assessment of outlier occurrence patterns, such as analyzing whether there are outlier patterns occurring at specific times each week. This information is extremely important for understanding the root causes of outliers, providing data support for further trend analysis and a scientific basis for decision-making, resulting in the outlier frequency results within the time window.

[0100] Based on the anomaly frequency results within the time window, the development and changes of the anomaly are assessed using the following formula:

[0101]

[0102] Calculate the outlier trend index R within the differential time window. b The time trend analysis results are obtained;

[0103] Where b represents the time variable, β0, β1, and γ are regression coefficients, ∈ b The term represents the random error, where α and δ are adjustment coefficients, and e is the natural constant.

[0104] By combining exponential and square root transformations, the formula improves the sensitivity to rapidly changing trends in data and the resistance to small random fluctuations, effectively balancing the response to abnormal data with the stability of the overall data trend. It is especially suitable for abnormal trend analysis in dynamically changing environments.

[0105] Example data given: b = 5 (days);

[0106] α = 1.2 (dynamically adjusted based on the variation of historical data);

[0107] β0 = 0.5 (baseline value obtained through historical data stability analysis);

[0108] β1 = 0.3 (based on regression analysis results of long-term data trends);

[0109] γ = 0.05 (trend acceleration factor identified from historical data);

[0110] δ = 1.5 (calculated based on the standard deviation of historical errors);

[0111] ∈ b =0.2 (mean error calculated from previous anomaly detection data);

[0112] Calculation process:

[0113] e γb =e 0.05×5 ≈1.284; This indicates the time growth factor.

[0114] Displays adjustments for random errors;

[0115] R b =1.2×0.5+0.3×1.284+1.5×0.447≈2.095;

[0116] The results indicate that the outlier trend index is 2.095 five days from now, which means that the trend of anomalies will be significantly enhanced during this period. Special monitoring and response measures need to be taken at this time to ensure the stability and security of the system.

[0117] Please see Figure 4 The specific steps for initializing the risk area are as follows:

[0118] Based on the time trend analysis results, time periods with data fluctuations exceeding the threshold are filtered out to identify potential risk events and obtain records of potential risk time periods.

[0119] By carefully analyzing patients' diabetes-related indicator data over a period of time, fluctuation characteristics in the data were identified, especially abnormal data points exceeding a predetermined threshold. This threshold was set based on the needs of disease management and mainly referenced daily blood glucose measurement changes, including statistical parameters such as the average, standard deviation, and coefficient of variation of blood glucose measurements. These statistical parameters were obtained through actual data monitoring. When determining the fluctuation range, not only the deviation of individual measurements was considered, but also the fluctuation trend of data over multiple consecutive days was assessed to more accurately capture the predicted risk points. After initial screening, the data will be used for cluster analysis to more meticulously identify and classify different risk patterns. Each risk period is confirmed by the actual situation of specific blood glucose measurements exceeding the threshold, ensuring the practical application value and accuracy of the analysis results and obtaining a record of potential risk period times.

[0120] Based on the records of potential risk time periods, the time periods are grouped, and differentiated risk patterns are identified to obtain a risk pattern group set.

[0121] A detailed analysis and grouping of time periods was conducted using clustering algorithms. This process involved multiple steps, including selecting a suitable clustering algorithm, specifically K-means clustering, a widely used technique suitable for processing large datasets. Before applying the clustering algorithm, data standardization was performed to eliminate the influence of different dimensions between indicators and ensure the accuracy of the clustering results. The main purpose of clustering is to group data with similar risk characteristics together. This was achieved by calculating the distance between each data point and the remaining points, using Euclidean distance because it has an intuitive advantage in expressing the absolute differences between points. Through iterative optimization of cluster centers, the focus of the analysis and the feasibility of the operation were ensured, resulting in a risk pattern grouping set.

[0122] Grouping sets by risk patterns, using the formula:

[0123]

[0124] Calculate the risk value R for each group. c If the risk value exceeds the threshold, it is marked to obtain the initial risk area;

[0125] Among them, w j It is the risk weight, x cjis the risk value of the j-th indicator in the c-th group, and m is the number of indicators;

[0126] Multiple risk indicators are combined using a weighted summation method to calculate the composite risk value for each risk pattern group, with a weighting factor w. j This reflects the importance of different indicators in risk assessment, making the results closer to the actual risk situation;

[0127] There is a group set up, which includes two indicators: blood glucose level and body mass index, with corresponding risk weights of 0.6 and 0.4. The average blood glucose level and body mass index in this group are 180 and 25, respectively.

[0128] The calculation process is as follows:

[0129] R c =0.6·180+0.4·25=108+10=118;

[0130] The results indicate that for this specific risk pattern group, the overall risk value is 118. This value can be directly used to assess the risk level of the group. Once the set risk threshold is exceeded, the group is marked as a high-risk area, which will directly affect subsequent disease management and intervention plans. This effectively identifies some risk pattern groups as having high risk and requiring management measures.

[0131] Please see Figure 5 The specific steps for obtaining risk assessment indicators are as follows:

[0132] Key data indicators for diabetic patients were extracted from the initial risk area, including blood glucose levels, blood pressure, and weight records, to obtain a set of key data indicators.

[0133] The initial risk zone delineation was achieved through detailed analysis of key data indicators, including crucial information such as blood glucose levels, blood pressure, and weight records for diabetic patients. Comprehensive data processing and analysis allowed for a more accurate identification of predicted health risk zones. This process involved data collection and processing, including data cleaning, where outlier removal was critical; for example, excessively high or low blood pressure values ​​were removed to ensure the quality and reliability of the dataset. Statistical analysis methods were used to calculate the mean and standard deviation of each data indicator. These statistics not only help understand the overall distribution of the data but also assess the contribution of each indicator to the overall risk assessment. A higher standard deviation in blood glucose data indicates instability in the patient's glycemic control, a significant factor in assessing the risk of diabetic patients, resulting in the set of key data indicators.

[0134] Statistical analysis is performed on the set of key data indicators to calculate the mean and standard deviation of differentiated key data indicators, assess the risk contribution of each indicator, and obtain the statistical analysis results.

[0135] When grouping risk patterns, the clustering algorithm used groups risk regions based on the similarity of data across different time periods. This process involves several key steps. The selection of the clustering algorithm and parameter settings are based on specific data characteristics, such as the number and type of clusters. Correct parameter selection is crucial for accurately identifying risk patterns. Each group of data reflects a specific risk pattern, which is achieved by analyzing the similarity and differences of data points within each group. By identifying similar patterns of elevated blood glucose over specific time periods, potential risks of diabetic complications can be identified. Further analysis includes assessing the prevalence of risk patterns across different patient populations and their potential impact on patient health. This helps healthcare providers develop targeted interventions to mitigate the impact of potential risks and yields statistical analysis results.

[0136] Using statistical analysis results, the criticality of indicators is measured based on data volatility, using the following formula:

[0137]

[0138] Calculate the coefficient of variation V for each indicator. k Risk assessment indicators are obtained;

[0139] Where, x k It represents a single data point, μ and σ are the mean and standard deviation of the indicator, respectively, and p is the number of indicators k;

[0140] Risk is measured by calculating the coefficient of variation of each indicator. A high coefficient of variation means that the data point fluctuates more relative to the mean, indicating a higher risk. Therefore, the formula can help to effectively identify and prioritize data indicators with high risk.

[0141] The mean was set to 100, the standard deviation to 15, the sample size to 50, and the data points were distributed as (90, 110, 95, 105, 115).

[0142] The coefficient of variation is calculated as follows:

[0143]

[0144] The results indicate that the given data indicators exhibit significant volatility, suggesting a high risk and highlighting the critical need for managing diabetes patients, as highly volatile blood glucose levels require more intensive monitoring and intervention.

[0145] Please see Figure 6 The specific steps for obtaining the health transition probability matrix are as follows:

[0146] Risk assessment indicators, including blood glucose control, weight change and diabetic complication records, were used. These indicators were standardized and converted into comparable scoring scales to obtain standardized risk scores.

[0147] By applying risk assessment indicators, including blood glucose control, weight change, and records of diabetic complications, data standardization is performed to transform these indicators into comparable scoring scales. This process involves multiple data processing steps, quantifying each indicator: blood glucose levels are quantified into a blood glucose control score, weight change is converted into a weight change score based on percentage change, and diabetic complications are classified and scored according to severity. The scores are then normalized to eliminate the influence of different scales, ensuring data consistency and comparability. This series of operations not only ensures data quality but also improves the accuracy of subsequent model predictions. The key to this process is ensuring that each risk indicator fairly reflects its influence in the assessment model, resulting in standardized risk scores.

[0148] By combining standardized risk scores with the health data of diabetic patients, the correlation is analyzed using the following formula:

[0149]

[0150] Estimate the transition probability P between different health states or Construct a health transition probability matrix;

[0151] Among them, S or α represents the standardized risk score indicating a transition from health state o to health state r. o β u and β r It is the adjustment coefficient, exp is the natural exponential function, and S ou A standardized risk score representing a transition from health state 0 to health state u;

[0152] By introducing a standardized risk score S from state o to r or and the adaptation parameter α o and β r This enhances the model's sensitivity to changes in health status and its adaptability to individual differences, helps to accurately predict the probability of health transition in diabetic patients, and provides a basis for developing personalized recovery recommendations;

[0153] There are three health states, A, B, and C, with corresponding standardized risk scores of S1, S2, and S3, respectively. oA =0.5, S oB =1.0, S oC =1.5; Adaptability parameter α o =0.2, βA =0.3, β B =0.5, β C =0.7;

[0154] Calculate the transition probability from state o to A using the formula:

[0155]

[0156] The results indicate that the transition probability from state 0 to state A is 25.5%, reflecting the probability of a patient transitioning from state 0 to state A under the current health score and parameter configuration.

[0157] Please see Figure 7 The specific steps for obtaining health status prediction results are as follows:

[0158] Using the health transition probability matrix, the applicability and consistency of predicting the health status of diabetic patients in the future time period are evaluated through data fit analysis of the health transition probability matrix, and fit score records are obtained.

[0159] By utilizing the constructed health transition probability matrix, a historical data fit analysis of the model is conducted to evaluate the model's applicability and accuracy in predicting future health status. By comparing the differences between the model's past predictions and actual health outcomes, statistical error analysis methods, such as mean squared error and coefficient of determination, are used to assess the accuracy of the predictions. This quantifies the model's adaptability, reflecting its reliability in processing future data. The more accurate the model is in predicting future patient health status, the better. This scoring mechanism not only helps medical professionals understand the model's performance but also ensures that the health prediction model can be effectively adjusted and optimized in practical applications, resulting in a fit score record.

[0160] Based on the fit score records and combined with the real-time health data of diabetic patients, the following formula is used:

[0161]

[0162] Calculate the probability prediction value of health status to obtain the health status prediction result;

[0163] Where P(t+1) is the predicted health status probability at the next time point, P(t) is the health status probability at the real-time time point, and x j (t) is the j-th indicator at time t, β j Here, B is the regression coefficient, B is the smoothing parameter, and m is the number of indicators.

[0164] The smoothing parameter B and the weighting coefficient β of the health indicators are used. jBy introducing this feature, the model can flexibly adjust its response to historical and new input data, improving the accuracy and adaptability of predictions. This approach enables the model to quickly adapt to recent changes in health data while maintaining the influence of historical data.

[0165] The health status probability P(t) at the current time point t is set to 0.7, and the smoothing parameter B is set to 0.3. This parameter reflects the weight the model assigns to new data while retaining the influence of historical data. Three health indicators are set, with corresponding coefficients β. j The values ​​are 0.5, 0.3, and 0.2 respectively, quantifying the contribution of each health indicator to the prediction model. The current value x of each health indicator... j (t) represents 1, 0, and 1 respectively, indicating specific health status data such as blood glucose control and weight change:

[0166] P(t+1)=(1-0.3)×0.7+0.3×(0.5×1+0.3×0+0.2×1)=0.49+0.3×(0.5+0+0.2)=0.49+0.3×0.7=0.49+0.2=0.70;

[0167] The results show that at the next time point t+1, the predicted probability of health status is 0.70, which represents the probability that the patient will maintain a stable health status. This result directly supports the output of the health status prediction, showing that the model can effectively maintain the accuracy of the prediction.

[0168] Please see Figure 8 The specific steps to obtain a health management plan for diabetic patients are as follows:

[0169] Using the health status prediction results, analyze the risk level and potential health problems of diabetic patients, identify key monitoring indicators and necessary intervention periods, and generate a draft health monitoring plan for diabetic patients.

[0170] An initial analysis of health status prediction results is conducted to assess the current health status of diabetic patients, including physiological indicators such as blood glucose levels, blood pressure, and heart rate, as well as lifestyle habits such as diet, activity levels, and sleep patterns. Data derived from patients' daily monitoring records is compared with their historical health data to identify new or exacerbating health risks. Based on the prediction results, acute or chronic health problems are identified, based on big data health analysis. For example, sudden fluctuations in blood glucose or persistently high blood pressure are important factors in determining monitoring indicators. Necessary intervention periods are determined based on risk factors, using the patient's past medical records and pattern recognition results to allow for proactive intervention in case of health problems. Such interventions are not only timely and effective but also tailored to individual patient needs, aiming to prevent potential problems from escalating. Adjusting the monitoring frequency is equally important. Monitoring frequency and intervention measures are adjusted based on real-time patient data and long-term health trends to maximize monitoring effectiveness, minimize patient discomfort, accurately reflect the patient's actual needs, and improve the timeliness and accuracy of monitoring, generating a draft health monitoring plan for diabetic patients.

[0171] Implement the draft health monitoring plan for diabetic patients, adjust the monitoring frequency, and adopt the following formula:

[0172] R adj =R base ·(1+α·P risk );

[0173] To assess the effectiveness and impact of the draft health monitoring plan for diabetic patients and to obtain a health management plan for diabetic patients;

[0174] Among them, R adj R represents the adjusted monitoring frequency. base This is the basic monitoring frequency, α is the adjustment coefficient, and P... risk It is a risk score for patients with diabetes;

[0175] Adjusting the frequency of health monitoring based on the patient's risk score makes the monitoring plan for diabetic patients more personalized and targeted. The monitoring frequency can be adjusted according to the patient's specific health condition, rather than using a one-size-fits-all approach, which helps to improve the effectiveness of monitoring and reduce unnecessary interventions.

[0176] Set the basic monitoring frequency R base Once a week, the patient's risk score P risk The value is 0.3, and the adjustment coefficient α is set to 0.5, which means that for each additional risk point, the monitoring frequency increases by 50%.

[0177] The adjusted monitoring frequency R was calculated. adj for:

[0178] R adj =1·(1+0.5·0.3)=1·1.15=1.15;

[0179] This means the monitoring frequency should be adjusted from once a week to 1.15 times a week. This result indicates that for patients with moderate health risks, increasing the monitoring frequency can better identify potential health problems and allow for timely intervention, improving the overall effectiveness of health management. This adjustment allows the diabetes patient monitoring program to be flexibly adjusted according to the patient's actual health condition, ensuring that patients receive timely attention and treatment in the face of any sudden or predicted health risks, achieving the goals of reducing the incidence of acute events and optimizing chronic disease management.

[0180] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A diabetes management system based on big data, characterized in that, The system includes: The anomaly monitoring module collects blood glucose, blood pressure, and weight data from diabetic patients, verifies the completeness of the data, removes unqualified data, identifies and analyzes anomalies in the data to obtain anomaly identification results, and performs time series analysis on the anomaly identification results to analyze the development trend of the abnormal data and obtain time trend analysis results. Based on the time trend analysis results, the risk analysis module analyzes the clustering characteristics of abnormal data, identifies the clustering distribution, generates an initial risk region, performs a risk assessment on the initial risk region, identifies the risk level of diabetes symptoms, and obtains risk assessment indicators. The health status simulation module uses the aforementioned risk assessment indicators to dynamically simulate the health status of diabetic patients, analyzes the differentiated health status transition probabilities, constructs a health transition probability matrix, uses the health transition probability matrix to predict the patient's health status in the future time period, evaluates the consistency of the prediction, and obtains the health status prediction results. Based on the health status prediction results, the monitoring and management module identifies key monitoring indicators and intervention time periods, optimizes health monitoring parameters, generates a draft health monitoring plan for diabetic patients, adjusts the monitoring frequency, evaluates the effectiveness and impact of the implementation of the draft health monitoring plan for diabetic patients, and obtains a health management plan for diabetic patients. The specific steps for obtaining the time trend analysis results are as follows: Using the anomaly identification results, extract the timestamp of the anomaly occurrence, record the time of the anomaly occurrence, track the time of the anomaly occurrence in real time, and obtain a list of timestamps for the anomaly values; A time window analysis was performed on the time-stamped list of outliers. By statistically analyzing the daily and weekly data, the frequency of outliers within the differentiated time window was evaluated, the outlier frequency was calculated, and the time window outlier frequency result was obtained. Based on the anomaly frequency results within the aforementioned time window, the development and changes of the anomaly are assessed using the following formula: ； Calculate the outlier trend index within the differential time window The time trend analysis results are obtained; in, Represents a time variable. , and For regression coefficients, For random error term, and To adjust the coefficient, It is a natural constant; The specific steps for obtaining the health transition probability matrix are as follows: Using the aforementioned risk assessment indicators, including blood glucose control, weight change, and records of diabetic complications, the risk assessment indicators are standardized and converted into comparable scoring scales to obtain standardized risk scores. By combining the standardized risk score with the health data of diabetic patients, the correlation is analyzed using the following formula: ； Estimate the transition probability between differentiated health states Construct a health transition probability matrix; in, Represents health status Transition to a healthy state Standardized risk scoring, , and It is an adjustment factor. It is a natural exponential function. Represents health status Transition to a healthy state Standardized risk scoring.

2. The diabetes management system based on big data according to claim 1, characterized in that, The specific steps for obtaining the anomaly identification results are as follows: Collect blood glucose, blood pressure, and weight data from diabetic patients, verify the integrity of the data, filter data that meets quality standards, and generate a quality-filtered dataset. Anomaly identification is performed on the dataset after quality screening. By calculating the weighted Euclidean distance between each data point and its neighboring data points and comparing it with a dynamically adjusted threshold, data points that deviate from the normal range are identified, and an anomaly data point set is generated. Time series analysis was performed on the aforementioned abnormal data point set to analyze the data development trend, according to the formula: ； Calculate the time-weighted moving average of the outliers to obtain the anomaly identification results; in, Representing a point in time abnormal trends Representing a point in time The One abnormal data point, Representing a point in time The Time weight of each data point This represents the number of outlier data points.

3. The diabetes management system based on big data according to claim 1, characterized in that, The specific steps for obtaining the initial risk region are as follows: Based on the time trend analysis results, time periods with data fluctuations exceeding the threshold are filtered out to identify potential risk events and obtain records of potential risk time periods. Based on the records of the potential risk time periods, the time periods are grouped, and risk patterns of different types are identified to obtain a risk pattern group set; Based on the risk pattern grouping set, the formula is used: ； Calculate the risk value for each group If the risk value exceeds the threshold, it is marked to obtain the initial risk area; in, It is a risk weight. It is the first The first indicator in the Group risk value, It refers to the number of indicators.

4. The diabetes management system based on big data according to claim 3, characterized in that, The specific steps for obtaining the risk assessment indicators are as follows: Key data indicators of diabetic patients are extracted from the initial risk area, including blood glucose levels, blood pressure and weight records of diabetic patients, to obtain a set of key data indicators. Statistical analysis is performed on the set of key data indicators to calculate the mean and standard deviation of the differentiated key data indicators, assess the risk contribution of each indicator, and obtain the statistical analysis results. Using the statistical analysis results, the criticality of the indicators is measured based on the volatility of the data, using the following formula: ； Calculate the coefficient of variation for each indicator. Risk assessment indicators are obtained; in, It is a single data point. and These are the mean and standard deviation of the indicator, respectively. It is an indicator The quantity.

5. The diabetes management system based on big data according to claim 1, characterized in that, The specific steps for obtaining the health status prediction results are as follows: Using the health transition probability matrix, the applicability and consistency of predicting the health status of diabetic patients in the future time period are evaluated through data fit analysis of the health transition probability matrix, and a fit score record is obtained. Based on the aforementioned fit score records, combined with the real-time health data of diabetic patients, the following formula is used: ； Calculate the probability prediction value of health status to obtain the health status prediction result; in, It is the probability of predicted health status at the next point in time. It represents the probability of health status at a real-time point in time. At a certain point in time The One indicator, It is the regression coefficient. It is a smoothing parameter. It refers to the number of indicators.

6. The diabetes management system based on big data according to claim 5, characterized in that, The specific steps for obtaining the health management plan for diabetic patients are as follows: Using the health status prediction results, analyze the risk level and potential health problems of diabetic patients, identify key monitoring indicators and necessary intervention periods, and generate a draft health monitoring plan for diabetic patients. Implement the aforementioned draft health monitoring plan for diabetic patients, adjust the monitoring frequency, and use the following formula: ； To assess the effectiveness and impact of the draft health monitoring plan for diabetic patients and to obtain a health management plan for diabetic patients; in, This represents the adjusted monitoring frequency. It is the basic monitoring frequency. It is an adjustment factor. It is a risk score for patients with diabetes.

Images