Intelligent evaluation device for health quality of diabetic patient based on machine learning

Abstract

The invention discloses a diabetic health attainment intelligent evaluation device based on machine learning, and the device comprises a Raspberry Pi 5 embedded platform, a touch display screen, a microphone, and a loudspeaker. A health attainment level prediction module, a user information collection and health attainment level evaluation module and a security and privacy protection module are integrated on the raspberry Pi 5 embedded platform. According to the invention, the logistic regression algorithm is adopted to carry out real-time analysis on the personal basic information of the user, and health attainment risk assessment results and personalized suggestions can be generated within several seconds. Compared with a traditional questionnaire survey evaluation method, the system remarkably improves the efficiency of health management. The system accurately reflects the health attainment level of the diabetic patient based on a prediction model constructed by multi-center real world data. Through actual measurement, the evaluation accuracy of the system is improved by about 25% compared with that of a traditional questionnaire-based evaluation method. By means of the improvement, evaluation errors are remarkably reduced, and more reliable health guidance is provided for patients.

Description

Technical Field

[0001] The present invention belongs to the intersection of artificial intelligence and medicine, and in particular relates to an intelligent health literacy assessor for diabetic patients based on machine learning. Background Art

[0002] Diabetes is a group of metabolic diseases characterized by chronic hyperglycemia. It is estimated that by 2040, the number of people aged 20-79 with diabetes will rise to 642 million, which will have a significant impact on social, financial and health systems. Health literacy (HL) refers to the ability to discover, understand and use health-related information or services, including basic literacy, self-efficacy, self-monitoring and management capabilities. The importance of health literacy has received widespread attention. At the scientific research level, the impact of health literacy on diabetic patients has been confirmed by many studies. Good health literacy can significantly enhance diabetic patients' ability to understand and deal with the disease, help them control their diet more effectively to reduce nutritional risks, thereby reducing the possibility of depression, reducing negative perceptions of the disease, and reducing the burden of treatment. Improving health literacy can significantly improve the disease management and quality of life of diabetic patients.

[0003] At present, the health literacy of diabetic patients is mainly evaluated by questionnaires. Through questionnaire surveys, the health literacy of diabetic patients is evaluated and the influencing factors are analyzed, especially to screen out diabetic patients with low health literacy and carry out key interventions, which can not only improve the quality of life of diabetic patients, but also reduce medical expenses and improve social and economic benefits.

[0004] At present, questionnaires, as the main means of evaluation, have some limitations: 1. Time-consuming: There are many items in the questionnaire survey, and this method takes a considerable amount of time, even up to several months, from organizing the survey, staffing, to collecting questionnaires and statistically analyzing data. 2. Low response rate: The response rate of the questionnaire survey may be low. Although it can be improved through some reward strategies, it will bring additional costs; 3. High cost: Because middle-aged and elderly people do not have sufficient Internet usage skills, paper questionnaires are indispensable, and the cost of paper surveys is high, mainly due to the expenses of printing, mailing and data entry. 4. There are certain requirements for users: self-evaluation questionnaires require patients to have a certain reading comprehension ability, and some patients with low cultural level still need to rely on the assistance of other people to complete; 5. Subjectivity of questionnaire items: The items of health literacy evaluation often contain subjective judgment questions of patients, which have a certain impact on the objectivity and accuracy of the evaluation results. In summary, it is necessary to develop a more efficient and convenient evaluation tool.

[0005] Machine learning is the core of artificial intelligence. It is a technology that uses statistical and computer science methods to enable computers to learn from data and continuously improve. Its application in the field of diabetes mainly includes the following two aspects: 1. Prediction of diabetes risk, such as using gradient enhancement machines to predict patients' future diabetes, and using logistic models to predict the risk of gestational diabetes. 2. Prediction of diabetes complications: such as using random forests to build a weighted risk model to predict the risk of diabetic foot, and using ensemble learning methods to build a risk prediction model for diabetic retinopathy. There is no application of machine learning in the prediction model of health literacy level of diabetic patients. Summary of the invention

[0006] Purpose of the invention: The purpose of the present invention is to provide an intelligent assessor for the health literacy of diabetic patients based on machine learning. The Delphi expert consultation method was used to construct an evaluation index system for the health literacy of diabetic patients, and a relevant questionnaire survey was conducted on diabetic patients in the community. Based on the survey data, the machine learning method was used to screen the important features that affect the health literacy level of diabetic patients, and a prediction model with a high accuracy rate was constructed. Using this assessor, the patient's health literacy level can be quickly predicted by inputting personal objective information such as age, income, educational background and lifestyle, without the need for patients to answer complex questionnaire questions. This is conducive to the timely discovery of diabetic patients with low health literacy, and provides guidance for the formulation of personalized patient intervention measures, thereby contributing to improving the patient's health level and promoting health.

[0007] Technical solution: The present invention provides a machine learning-based intelligent health literacy assessor for diabetic patients, comprising a housing, a Raspberry Pi 5 embedded platform, a touch screen, a microphone and a speaker, wherein the Raspberry Pi 5 embedded platform is integrated with a health literacy level prediction module, a user information collection and health literacy level assessment module and a security and privacy protection module;

[0008] The health literacy level prediction module adopts the Delphi expert consultation method to construct a health literacy evaluation index system for diabetic patients, conducts questionnaire surveys on diabetic patients in hospitals and communities, and counts and analyzes the basic information of patients in the survey data and their scores in four aspects: health awareness, health knowledge, health behavior and health skills; uses the RFE algorithm to fit the model with demographic characteristics, obtains the importance ranking of all features, and then loops, removing the least important features each time, and finally determines the number of features of the best performing model, and uses the selected features to fit the best model; finally, based on the eight recommended features given by the RFE algorithm and the patient data collected in hospitals and communities, a logistic regression algorithm is used to construct an AI model for predicting the health literacy level of patients;

[0009] The user information collection and health literacy level assessment module includes a hardware layer, a software layer, a data collection and processing layer, and an assessment and suggestion layer;

[0010] The security and privacy protection module includes a data security and access control module and a compliance assurance and privacy protection module.

[0011] Furthermore, the eight recommended characteristics include age, education level, whether diet is controlled, whether exercise is controlled, the number of people living together, whether blood sugar is controlled by injection, duration of diabetes and income level.

[0012] Furthermore, the hardware layer: the system is based on the Raspberry Pi 5 embedded platform, integrating a touch display, a microphone and a speaker for user interaction and voice input and output, and providing a variety of external interfaces, including micro-HDMI*2, USB-C power interface, USB 2.0 and USB 3.0 interfaces, to support a variety of device connections and expansions;

[0013] The software layer: uses Raspberry Pi operating system Raspbian; user interface development uses Python and Tkinter library to build the user interaction interface; speech recognition integrates iFlytek API for speech-to-text processing; data processing and machine learning use Python programming language, and the main libraries include Pandas, scikit-learn and NumPy.

[0014] The data collection and processing layer: the system collects objective basic information of users and eliminates the difference in feature value scale through standardization processing StandardScaler;

[0015] The evaluation and recommendation layer: generates the patient's diabetes health literacy assessment results through the AI ​​model based on the input personal basic information. Based on the assessment results, the system generates personalized health management recommendations covering diet and exercise to achieve personalized health management.

[0016] Furthermore, the data security and access control module adopts a role-based permission management system to ensure that only authorized personnel can access and process sensitive data; the data is encrypted during transmission and storage; the compliance assurance and privacy protection module complies with privacy protection regulations GDPR and HIPAA to ensure the security and compliance of user data when providing health management and advice.

[0017] Furthermore, the health literacy level prediction module uses a logistic regression algorithm to construct an AI model for predicting the health literacy level of the target. Logistic regression maps the output of the linear model to the interval [0,1] through a Sigmoid function, thereby indicating the probability of insufficient health literacy in diabetic patients.

[0018] Define the linear model: Construct a linear model z = w1 age + w2 education level + ... + w8 income level + b, where w1 to w8 are weight coefficients and b is the bias option;

[0019] Apply the Sigmoid function: Pass the output z of the linear model through the Sigmoid function Mapped to the interval [0,1], we get the probability of poor health literacy among diabetic patients;

[0020] Sigmoid function:

[0021]

[0022] Model training: Use optimization algorithms such as maximum likelihood estimation or gradient descent to iteratively update weights w and biases b so that the probability distribution predicted by the model is close to the distribution of actual data. Suppose there are m training samples, each with a feature vector x_i, a corresponding label y_i (0 or 1), a weight vector w, a bias b, and a learning rate a. The update rule of gradient descent is expressed as:

[0023]

[0024] in, is the hypothesis function of logistic regression, is the jth feature of the i-th sample.

[0025] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0026] The intelligent assessment system for health literacy of diabetic patients of the present invention realizes efficient assessment of health literacy level of diabetic patients and generation of personalized health management suggestions through deep integration of embedded platform and artificial intelligence technology, which has significant advantages over the existing technology. The specific effects are as follows:

[0027] Improved evaluation efficiency: The present invention uses a logistic regression algorithm to analyze the user's basic personal information in real time, and can generate health literacy risk assessment results and personalized recommendations within a few seconds. Compared with traditional questionnaire assessment methods, this system significantly improves the efficiency of health management.

[0028] Enhanced assessment accuracy: The system uses a prediction model based on multi-center real-world data to accurately reflect the health literacy level of diabetic patients. After actual testing, the system's assessment accuracy is about 25% higher than the traditional questionnaire-based assessment method. This improvement significantly reduces assessment errors and provides patients with more reliable health guidance.

[0029] Improve user experience: By integrating voice recognition technology, users can directly input basic personal information by voice, which is more convenient to use. The touch screen provides an intuitive user interface and simplifies the operation steps. User satisfaction surveys show that the ease of use of this system has been widely recognized.

[0030] Data security and privacy protection: The system's built-in data encryption and permission management modules ensure the security and privacy compliance of user data. Compared with the looser privacy protection measures in the prior art, the data security mechanism of the present invention reduces the risk of user data leakage by about 60%, enhancing users' trust in the system.

[0031] Wide applicability: The modular design of the present invention allows it to be flexibly applied to different medical scenarios, including community health centers, hospital endocrinology departments, and personal family health management. Tests have shown that the system can operate stably in different scenarios without the need for complex equipment adjustments, greatly improving the applicability and popularity of the system.

[0032] Practical application effect verification: In the actual application of community health management, the present invention effectively helps medical workers screen out diabetic patients with low health literacy levels and provides scientific health management suggestions for these patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is the hardware diagram of the Raspberry Pi;

[0034] Figure 2 This is the system architecture diagram;

[0035] Figure 3 The following is a schematic diagram of the workflow;

[0036] Figure 4 It is a structural and component diagram of the present invention;

[0037] Figure 5 is the structural dimensions (front);

[0038] Figure 6 is the structural dimensions (side);

[0039] Figure 7 Start interface for users;

[0040] Figure 8 Assessment results and personalized recommendation interface for health literacy; DETAILED DESCRIPTION

[0041] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0042] The present invention proposes an intelligent assessor for the health literacy of diabetic patients based on machine learning and real-world data. It is based on the survey data on the health literacy levels of diabetic patients in hospitals and communities, and uses artificial intelligence technology to construct a predictive model for the health literacy levels of diabetic patients based on basic personal information. The aim is to collect objective personal information such as the patient's age, income, educational background and lifestyle through a simple question-and-answer method, so as to assess the patient's health literacy level. The invention is particularly suitable for quickly evaluating the health literacy level of diabetic patients. The AI ​​predictive model constructed using a machine learning algorithm screens out patients with low health literacy levels, which not only improves the accuracy and efficiency of disease management, but also provides a reference for personalized health interventions for patients. The specific technical means include the following modules:

[0043] Construct a prediction model for the health literacy level of diabetic patients:

[0044] Questionnaire survey: A scientific and reliable health literacy evaluation index system for diabetic patients was constructed using the Delphi expert consultation method. Relevant questionnaire surveys were conducted on diabetic patients in hospitals and communities. The basic information of patients in the survey data and their scores in four aspects, including health awareness, health knowledge, health behavior and health skills, were counted and analyzed to provide real-world data for building a predictive model.

[0045] Model construction: The RFE algorithm was used to fit the model to all relevant demographic characteristics, and the importance ranking of all features was obtained. Then the cycle was repeated, and the least important features were removed each time. Finally, the number of features of the best performing model was determined, and the selected features were used to fit the best model. In the end, the algorithm gave 8 recommended features: age, education level, whether diet control, whether exercise control, number of people living together, whether injections of drugs were used to control blood sugar, duration of diabetes, and income level. According to the 8 features given by the RFE algorithm, based on patient data collected from hospitals and communities, an AI model was constructed using a logistic regression algorithm to predict patients' health literacy levels.

[0046] User information collection and health literacy level assessment module:

[0047] Hardware layer: The system is based on the Raspberry Pi 5 embedded platform and integrates a 10-inch touch screen, microphone and speaker for user interaction and voice input and output. It provides a variety of external interfaces, including micro-HDMI*2, USB-C power interface, USB 2.0 and USB 3.0 interfaces, to support a variety of device connections and expansions. Figure 1 shown.

[0048] Software layer: Raspberry Pi operating system: Raspbian; User interface development: Use Python and Tkinter library to build a user-friendly interactive interface; Speech recognition: Integrate iFlytek API for speech-to-text processing; Data processing and machine learning: Use Python programming language, and the main libraries include Pandas, scikit-learn, NumPy, etc.

[0049] Data collection and processing: The system collects objective basic information of users (such as age, income, educational background, and lifestyle, etc.), and eliminates the scale differences of feature values ​​through standardization (using StandardScaler) to provide consistent data for model input.

[0050] Assessment and recommendations: The system generates the patient's diabetes health literacy assessment results through the AI ​​model based on the input personal basic information. Based on the assessment results, the system generates personalized health management recommendations, covering intervention measures such as diet, exercise and drug treatment, to help patients achieve personalized health management.

[0051] Security and privacy protection module:

[0052] Data security and access control: The system uses a role-based permission management system to ensure that only authorized personnel can access and process sensitive data, thereby ensuring the security of data access. Data is encrypted during transmission and storage to ensure the security of patient data.

[0053] Compliance assurance and privacy protection: The system strictly complies with relevant privacy protection regulations (such as GDPR, HIPAA) to ensure the security and compliance of user data when providing health management and disease prevention advice. Provide access rights to medical workers, allowing them to view health assessment results for the development of health management plans.

[0054] Working principle:

[0055] The structure and components of the invention, the intelligent evaluation system for health literacy of diabetic patients of the present invention is as follows Figure 4 As shown: It mainly includes the following components:

[0056] Embedded platform (①): Raspberry Pi 5 is used as the core processing unit, responsible for data acquisition, processing and overall control of the system.

[0057] External interface (②): Provides micro-HDMI*2, USB-C power interface, USB 2.0*2, USB 3.0*2 and Ethernet interface for device connection and data transmission.

[0058] Display and interaction device (③): includes a 10-inch touch screen for user interface display and user interaction.

[0059] Audio device (④): integrated with microphone and speaker for voice input and output, realizing voice interaction with users.

[0060] Data processing module (⑤): Use Python’s Pandas and NumPy libraries to perform data cleaning, standardization (using StandardScaler), and feature extraction to provide consistent input data for the prediction model.

[0061] Speech recognition module (⑥): Integrates iFlytek API to process user voice input and converts the voice into text data for further analysis by the system.

[0062] Security and privacy protection module (⑦): includes data encryption and role permission management to ensure the security and privacy compliance of user data.

[0063] Health literacy level prediction module (⑧): Based on the patient data collected from hospitals and communities, an AI model is constructed using a logistic regression algorithm to predict the patient's health literacy level. Logistic regression is a statistical learning method widely used in classification problems, especially binary classification problems. It maps the output of the linear model to the [0,1] interval through the Sigmoid function, thereby indicating the probability of insufficient health literacy in diabetic patients.

[0064] Define the linear model: Construct a linear model z = w1 age + w2 education level + ... + w8 income level + b, where w1 to w8 are weight coefficients and b is the bias term.

[0065] Apply the Sigmoid function: convert the output of the linear model Through the Sigmoid function Mapped to the interval [0,1], we get the probability of poor health literacy among diabetic patients;

[0066] Sigmoid function:

[0067]

[0068] Model training: Use optimization algorithms such as maximum likelihood estimation or gradient descent to iteratively update weights w and biases b so that the probability distribution predicted by the model is close to the distribution of actual data. Suppose there are m training samples, each with a feature vector x_i, a corresponding label y_i (0 or 1), a weight vector w, a bias b, and a learning rate a. The update rule of gradient descent is expressed as:

[0069]

[0070] in, is the hypothesis function of logistic regression, is the jth feature of the i-th sample.

[0071] Health management advice module (⑨): Generates personalized health management advice based on the user's health literacy level prediction results, including diet, exercise and drug treatment plans.

[0072] The relationship between the components:

[0073] The embedded platform (①) is the core control unit of the system, connected to the display and interaction device (③), responsible for displaying the interface and receiving user input.

[0074] The audio device (④) is connected to the embedded platform (①) through an interface to provide a voice interaction function.

[0075] The speech recognition module (⑥) is integrated with the embedded platform (①) to realize the recognition and processing of speech input, and the converted text data is further analyzed by the data processing module (⑤).

[0076] The data processing module (⑤) standardizes the collected data and provides it as input to the health literacy prediction module (⑦).

[0077] The health literacy level prediction module (⑧) analyzes the user's basic personal information through a model, generates a health literacy level assessment result, and sends the result to the health management suggestion module (⑨).

[0078] The health management suggestion module (⑨) generates personalized suggestions based on the evaluation results, and these suggestions are presented to the user through the display and interaction device (③).

[0079] The security and privacy protection module (⑦5) runs through the entire system to ensure data security and user privacy protection at every stage.

[0080] Mode of operation of the invention:

[0081] Step 1: User information collection

[0082] The user provides basic personal characteristics required for prediction, such as age, income, education background, and lifestyle, through a touch screen (③) or voice input device (④). The system uses a speech recognition module (⑥) to convert the voice input into text form.

[0083] Step 2: Data Preprocessing

[0084] The collected personal information is cleaned and standardized by the data processing module (⑤) to eliminate feature scale differences and ensure the consistency and accuracy of the model input.

[0085] Step 3: Health literacy level prediction model processing

[0086] The processed data is input into the health literacy level prediction module (⑧), and the system uses a logistic regression model to analyze the data and generate the patient's health literacy risk assessment results.

[0087] Step 4: Personalized health recommendations

[0088] The health management suggestion module (⑨) generates personalized health management suggestions for users based on the predicted health literacy level evaluation results, including diet, exercise and medication plans. These suggestions are presented through display devices (③) or fed back to users through voice devices (④).

[0089] Step 5: Data security and privacy protection

[0090] During the entire operation process, the system encrypts data through the security and privacy protection module (⑦) and uses role-based permission management to control access, ensuring the security and compliance of patient information.

[0091] Unique features of the present invention:

[0092] Innovative data processing and analysis: This paper uses logistic regression models and real-world data to construct a health literacy level prediction model based on personal objective basic information, which can achieve rapid and accurate prediction of the health literacy of diabetic patients. Compared with the currently commonly used questionnaire survey evaluation method, it can significantly improve the evaluation speed and the degree of personalization of health management recommendations.

[0093] More objective and accurate evaluation method: The present invention effectively avoids the subjective judgment in the traditional questionnaire evaluation process by analyzing and predicting the objective basic information entered by the patient, thereby improving the objectivity and accuracy of the evaluation results.

[0094] Efficient user interaction method: The system integrates text and voice input methods, which improves the user experience and makes the health assessment process more convenient.

[0095] Efficient data collection and real-time feedback mechanism: The system can be used for medical activities and personal health management, helping medical workers to quickly assess patients' health literacy levels and provide scientific support for personalized health interventions, thereby improving the efficiency and effectiveness of disease management.

[0096] Structural features

[0097] 1. Components

[0098] The intelligent evaluation system for health literacy of diabetic patients of the present invention includes the following main components:

[0099] Embedded platform (①): Raspberry Pi 5 as the core processing unit.

[0100] External interface (②): including micro-HDMI*2, USB-C power interface, USB 2.0*2, USB 3.0*2, Ethernet interface.

[0101] Display and interactive equipment (③): 10-inch touch screen.

[0102] Audio device (④): integrated microphone and speaker.

[0103] Data processing module (⑤): based on Python’s Pandas and NumPy libraries.

[0104] Speech recognition module (⑥): integrated with iFlytek API.

[0105] Security and privacy protection module (⑦): includes data encryption, role permission management and anonymized data access.

[0106] Health literacy prediction module (⑧): Based on the Delphi expert consultation method, a scientific and reliable health literacy evaluation index system for diabetic patients was constructed, combined with an AI model constructed using a logistic regression algorithm.

[0107] Health management advice module (⑨): used to generate personalized health advice.

[0108] 2. Construction of each component

[0109] Embedded platform (①):

[0110] Shape and size: The Raspberry Pi 5 is a rectangular board with a size of approximately 85.6mm x 56.5mm, which is compact and easy to install.

[0111] Materials: Mainly composed of high-density circuit boards, electronic components and metal connectors to ensure performance stability and durability.

[0112] Function: As the core processing unit, it is responsible for data collection, processing and overall control of the system.

[0113] External interface (②):

[0114] Type and location: Includes micro-HDMI*2 interface, USB-C power interface, USB 2.0*2, USB 3.0*2, Ethernet interface, located on the side of the embedded platform.

[0115] Material: Metal connector and plastic shell to ensure the stability and durability of the connection.

[0116] Function: Supports diverse device connections and expansions, such as external display, power supply, data transmission, and network connection.

[0117] Display and interactive equipment (③):

[0118] Shape and size: 10-inch touch display, moderate size, easy to place on the desktop and user operation.

[0119] Material: Adopt high-resolution LCD screen and solid glass panel, providing clear display and durable touch function.

[0120] Function: Display the user interface and provide touch input function for users to input data and interact.

[0121] Audio equipment (④):

[0122] Shape and size: Small built-in microphone and speaker, integrated around the display device.

[0123] Materials: The microphone is a combination of metal and plastic, and the speaker is a plastic shell with a built-in voice coil.

[0124] Function: The microphone is used for voice input and the speaker is used for voice output to achieve audio interaction with the user.

[0125] Data processing module (⑤):

[0126] Construction: Software module running on embedded platform, using Pandas and NumPy libraries for data processing.

[0127] Function: Clean and standardize the data input by users to provide high-quality data input for subsequent AI models.

[0128] Speech Recognition Module (⑥):

[0129] Structure: Through software integration of iFlytek API, it has no physical shape and is embedded in the system software architecture.

[0130] Function: Process user voice input and convert voice into text data.

[0131] Security and Privacy Protection Module (⑦):

[0132] Construction: Software module integrating data encryption, role-based permission management and anonymization processing.

[0133] Function: Ensure the security and privacy protection of user data during collection, processing and storage.

[0134] Health literacy prediction module (⑧):

[0135] Construction: A software module built on the basis of logistic regression algorithm, integrated into the system.

[0136] Function: Analyze the input and processed user standardized data to generate the patient's health literacy level assessment results.

[0137] Health management advice module (⑨):

[0138] Structure: A software module that works with the health literacy level prediction module to generate personalized health management recommendations based on the prediction results.

[0139] Function: Output personalized recommendations for diet, exercise, and medication management.

[0140] 3. Connections and relationships between components

[0141] Connection method: Each module is mainly connected to each other through the hardware interface and internal data bus of the embedded platform (①). The speech recognition module (⑥), data processing module (⑤), health literacy prediction module (⑧), health management suggestion module (⑨) and security and privacy protection module (⑦) are all software modules on the embedded platform, working together through data flow and logical connection.

[0142] Relative position: the display and interactive device (③) is connected to the display output interface of the embedded platform (①), the audio device (④) is connected to the embedded platform (①) in parallel with the display output interface, and the external interface (②) is located at the outer edge of the embedded platform (①) to facilitate the connection of other devices.

[0143] The embodiment of the present invention comprises the following steps:

[0144] Step 1: Device Setup and Initialization

[0145] First, fix the embedded platform Raspberry Pi 5 (①) in the device housing to ensure that it is stable. Raspberry Pi 5 is connected to the power supply through the USB-C interface (②) to provide power support for the system.

[0146] A 10-inch touch screen (③) is installed on the front of the device and connected to the Raspberry Pi 5 via a micro-HDMI port (②) to facilitate user interaction.

[0147] The microphone and speaker (④) are integrated above and below the display screen and connected to the embedded platform through an audio interface to realize voice input and output functions.

[0148] Step 2: Data collection and preprocessing

[0149] Users input basic personal information such as age, income, education background and lifestyle through the touch screen or voice. The voice input is processed by the voice recognition module (⑥) to convert the voice into text.

[0150] The collected data are standardized by the data processing module (⑤) and the StandardScaler is used for data normalization to eliminate the scale differences of the eigenvalues ​​and provide consistent data input for subsequent analysis.

[0151] Step 3: Health literacy level assessment

[0152] The processed standardized data is input into the health literacy prediction module (⑧). This module uses a logistic regression algorithm to analyze and evaluate the user's health literacy level and generate a personalized health literacy evaluation.

[0153] Step 4: Generate Personalized Recommendations

[0154] Based on the health literacy risk assessment results, the health management advice module (⑨) generates personalized health advice including diet, exercise and medication management. These advice are presented to the user through a display screen or voice output.

[0155] Step 5: Data security and privacy protection

[0156] During the data collection, processing and storage process, the security and privacy protection module (⑦) ensures the security and privacy compliance of user data through data encryption, role permission management and anonymization.

[0157] Material selection and composition

[0158] Core components: Raspberry Pi 5 embedded platform, 10-inch touch display, microphone and speaker.

[0159] Materials: Raspberry Pi is made of high-density circuit boards and electronic components; the display uses a high-resolution LCD screen and a solid glass panel.

Claims

1. An intelligent health literacy assessor for diabetic patients based on machine learning, characterized in that: It includes a housing, a Raspberry Pi 5 embedded platform, a touch display screen, a microphone and a speaker, wherein the Raspberry Pi 5 embedded platform is integrated with a health literacy level prediction module, a user information collection and health literacy level assessment module and a security and privacy protection module; The health literacy level prediction module adopts the Delphi expert consultation method to construct a health literacy evaluation index system for diabetic patients, conducts a questionnaire survey on diabetic patients in hospitals and communities, and collects and analyzes the basic information of patients in the survey data and their scores in four aspects: health awareness, health knowledge, health behavior and health skills; The RFE algorithm was used to fit the model to the demographic characteristics, and the importance ranking of all features was obtained. Then, the model was looped and the least important features were removed each time. Finally, the number of features of the best performing model was determined, and the selected features were used to fit the best model. Finally, based on the eight recommended features given by the RFE algorithm and the patient data collected from hospitals and communities, an AI model was constructed using a logistic regression algorithm to predict the patient's health literacy level. The user information collection and health literacy level assessment module includes a hardware layer, a software layer, a data collection and processing layer, and an assessment and suggestion layer; The security and privacy protection module includes a data security and access control module and a compliance assurance and privacy protection module.

2. According to claim 1, a machine learning-based intelligent health literacy assessor for diabetic patients, characterized in that: The eight recommended characteristics include age, education level, whether diet is controlled, whether exercise is controlled, the number of people living together, whether injection of drugs is used to control blood sugar, duration of diabetes and income level.

3. According to claim 1, a machine learning-based intelligent health literacy assessor for diabetic patients, characterized in that: The hardware layer: The system is based on the Raspberry Pi 5 embedded platform, integrating a touch screen, microphone and speaker for user interaction and voice input and output, and provides a variety of external interfaces, including micro-HDMI*2, USB-C power interface, USB2.0 and USB 3.0 interfaces, to support a variety of device connections and expansions; The software layer: uses Raspberry Pi operating system Raspbian; User interface development uses Python and Tkinter libraries to build user interaction interfaces; speech recognition integrates iFlytek API for speech-to-text processing; data processing and machine learning use Python programming language, and the main libraries include Pandas, scikit-learn, and NumPy; The data collection and processing layer: the system collects objective basic information of users and eliminates the difference in feature value scale through standardization processing StandardScaler; The evaluation and recommendation layer: generates the patient's diabetes health literacy assessment results through the AI ​​model based on the input personal basic information. Based on the assessment results, the system generates personalized health management recommendations covering diet and exercise to achieve personalized health management.

4. According to claim 1, a machine learning-based intelligent health literacy assessor for diabetic patients, characterized in that: The data security and access control module adopts a role-based permission management system to ensure that only authorized personnel can access and process sensitive data; data is encrypted during transmission and storage; the compliance assurance and privacy protection module complies with privacy protection regulations GDPR and HIPAA to ensure the security and compliance of user data when providing health management and advice.

5. According to claim 1, a machine learning-based intelligent health literacy assessor for diabetic patients, characterized in that: The health literacy level prediction module uses a logistic regression algorithm to construct an AI model for predicting the health literacy level of the target. Logistic regression maps the output of the linear model to the interval [0,1] through the Sigmoid function, thereby indicating the probability of insufficient health literacy in diabetic patients. Define the linear model: Construct a linear model z = w1 age + w2 education level + ... + w8 income level + b, where w1 to w8 are weight coefficients and b is the bias option; Apply the Sigmoid function: convert the output of the linear model Through the Sigmoid function Mapped to the interval [0,1], we get the probability of poor health literacy among diabetic patients; Sigmoid function: Model training: Use optimization algorithms such as maximum likelihood estimation or gradient descent to iteratively update weights w and biases b so that the probability distribution predicted by the model is close to the distribution of actual data. Suppose there are m training samples, each with a feature vector x_i, a corresponding label y_i, a weight vector w, a bias b, and a learning rate a. The update rule of gradient descent is expressed as: in, is the hypothesis function of logistic regression, is the jth feature of the i-th sample.

Images