Personalized medical equipment configuration and management system based on digital twinning technology

Abstract

The invention discloses a personalized medical equipment configuration and management system based on a digital twinning technology. The personalized medical equipment configuration and management system comprises a patient demand analysis module for extracting patient personalized demands and group common demands; the equipment state monitoring module is used for collecting equipment operation data in real time and uploading the data to the cloud for analysis; the digital twinborn model construction module is used for creating an equipment virtual model and synchronizing a physical state and behavior logic; and the optimization decision and configuration module is used for generating an optimal equipment configuration scheme based on data driving. In the invention, through personalized management and dynamic configuration, parameters are flexibly adjusted according to scenes, the use efficiency of equipment is maximized, and a foundation is laid for subsequent management; on the basis of deep analysis of equipment operation data, full-life-cycle management is realized by means of real-time monitoring and automatic configuration optimization, the effective service time of equipment is prolonged, and a'monitoring-optimization 'closed loop is formed; a three-dimensional digital twinborn model is introduced, the equipment state, the health condition and configuration suggestions are visually displayed, and an interactive interface is created.

Description

Technical Field

[0001] The present invention belongs to the field of medical equipment management technology, specifically to a personalized medical equipment configuration and management system based on digital twin technology. Background Art

[0002] In the field of modern medical device management, the demand for personalized medical device configuration and management is steadily increasing. Different medical institutions and departments have varying requirements for device performance, operating environments, and maintenance. Traditional medical device management methods rely heavily on manual troubleshooting and scheduled maintenance. This approach is inefficient given the large variety of devices and the heavy workload, and it lacks accurate, real-time management and configuration of devices. Furthermore, traditional methods lack comprehensive analysis of the usage status and historical data of different devices, making it difficult to achieve data-driven configuration optimization and predictive maintenance.

[0003] Existing equipment management solutions primarily rely on Internet of Things (IoT) platforms based on basic database systems and sensor networks, enabling daily equipment scheduling and status monitoring through centralized management. Some advanced management systems incorporate basic artificial intelligence (AI) analysis capabilities to predict equipment failures or maintenance needs. While existing IoT systems can partially achieve remote management through data collection and monitoring, these systems suffer from significant deficiencies in personalized management, real-time optimization, and intelligent decision-making, making them unable to flexibly address a wide range of complex medical scenarios. Summary of the Invention

[0004] The purpose of the present invention is to provide a personalized medical equipment configuration and management system based on digital twin technology to solve the problems in the existing technology proposed in the background technology, such as the lack of real-time personalized adjustment capabilities, insufficient data utilization, difficulty in achieving comprehensive equipment life cycle management, and lack of three-dimensional visualization support based on digital twins.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] Personalized medical device configuration and management system based on digital twin technology, including:

[0007] Patient demand analysis module: extracting patients’ individual needs and group common needs;

[0008] Equipment status monitoring module: collects equipment operation data in real time and uploads it to the cloud for analysis;

[0009] Digital twin model building module: creates a virtual model of the device, synchronizing the physical state and behavioral logic;

[0010] Optimization decision and configuration module: Generate optimal equipment configuration solutions based on data-driven.

[0011] According to the above technical solution, the patient needs analysis module includes:

[0012] Data acquisition: Extracting basic patient information from the database, including patient medical records, feedback, questionnaires, and doctor's notes, and using NLP to extract high-frequency keywords in patient feedback;

[0013] Sentiment analysis: Use NLP methods to perform sentiment analysis on patient feedback to obtain sentiment labels and sentiment scores;

[0014] Generate patient demand reports: Based on the previous sentiment analysis results, summarize the patient's specific needs into a demand report, identify high-frequency demand combinations through the Apriori algorithm, and generate individual and group demand reports.

[0015] According to the above technical solution, NLP is used to extract high-frequency keywords from patient feedback:

[0016] TF-IDF(t,d)=TF(t,d)×IDF(r)

[0017] Where TF(t,d) is the frequency of term t in document d, and IDF(t) is the inverse document frequency of term t.

[0018] According to the above technical solution, NLP methods are used to perform sentiment analysis on patient feedback, and the sentiment labels and sentiment scores are obtained as follows:

[0019] Data preprocessing: Clean and preprocess text data to remove stop words and punctuation, unify text case, and extract stems;

[0020] Sentiment Analysis Model: Use the BERT model to perform sentiment analysis on patient feedback. We assign sentiment labels, representing different sentiment directions by setting positive, neutral, and negative. The processed patient feedback is input into the BERT model, which then outputs the probability distribution of each sentiment label.

[0021] The emotion category is selected based on the maximum probability, as shown in the following formula:

[0022] Sentiment=aregmax(BERT(f patient ))

[0023] Among them, f patient For patient feedback text, the BRET model outputs a probability distribution of sentiment labels; the patient's emotions are further quantified through sentiment scores (a numerical value ranging from -1 to 1).

[0024] According to the above technical solution, the Apriori algorithm identifies high-frequency demand combinations and generates individual and group demand reports as follows:

[0025] Step S1, constructing a patient demand dataset; converting each patient's demand into a set of items and constructing a database, where each record represents a patient's demand set;

[0026] In step S2, the Apriori algorithm is used to extract the common needs of the patient group. Specifically, the Apriori algorithm is used to traverse all patient need sets and calculate the support of each need item. The minimum support threshold is set to filter out the item sets that meet the conditions, as shown in the following formula:

[0027]

[0028] Where, the numerator (Number of transactions containing X) represents the number of transactions (such as patient demand records) containing demand item X; the denominator (Total number of transactions) represents the total number of transactions;

[0029] Formula meaning: Support (Support X) is the frequency of occurrence of requirement item X in all transactions, that is, the proportion of transactions containing X to the total number of transactions. For example, if there are 100 transactions in total, 30 of which contain X, then Indicates that X appears in 30% of transactions; by setting a minimum support threshold (such as 0.2), the set of demand items that appear frequently can be screened out (when 0.3 ≥ 0.2, X meets the condition), providing a basis for the Apriori algorithm to mine common needs of patients.

[0030] Step S3: Combine the frequent itemsets from the previous round to generate new candidate itemsets, calculate their support, and continue to screen itemsets whose support meets the threshold;

[0031] Step S4, repeat the above process until no new frequent item sets can be found; the frequent item sets finally generated are the common needs of the patient group.

[0032] According to the above technical solution, equipment status monitoring includes:

[0033] Real-time data collection: Obtain equipment operating parameters through sensors, including equipment temperature, vibration, and usage time;

[0034] Data transmission and storage: Upload data to the cloud time series database via IoT protocols (MQTT / HTTP);

[0035] Anomaly detection: Use the moving average method to identify abnormal equipment status and trigger an early warning.

[0036] According to the above technical solution, the moving average method is used to identify abnormal equipment status and trigger an early warning as follows:

[0037]

[0038] Where, MA t is the moving average at time t, x i is the i-th data point in the data sequence. If the sensor data exceeds the set threshold range, an alarm signal is triggered.

[0039] According to the above technical solution, the construction of the digital twin model includes:

[0040] Physical and behavioral modeling: Define the physical properties and behavioral logic of the device. Physical properties include device size and material. Behavioral logic includes differential equations for degree change and state transition conditions.

[0041] Real-time data fusion: Integrate sensor data through Kalman filter and dynamically update virtual model status;

[0042] 3D visualization: Use 3D tools to display the real-time status of equipment and support interactive monitoring.

[0043] According to the above technical solution, the optimization decision and configuration include:

[0044] Data input: Collect device usage and patient needs data; integrate device usage data with patient needs level;

[0045] Intelligent decision-making: The configuration model is trained through the decision tree algorithm, and key features are selected based on information gain;

[0046] Solution output: Generate dynamic configuration suggestions and verify their rationality through simulation.

[0047] According to the above technical solution, the configuration model is trained by the decision tree algorithm as follows:

[0048] First, we select the key features that affect device configuration. Then, we train a decision tree based on these key features to determine the configuration rules, as shown in the following formula:

[0049]

[0050] Where H(D) is the entropy of the data set D, D v The essence of information gain InfoGain(D,A) is: the uncertainty H(D) of the original data set D and the average uncertainty of each subset after being divided by feature A. The larger the difference, the greater the contribution of feature A to reducing data uncertainty, and the more suitable it is as a partitioning feature.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] In the present invention, personalized management and dynamic configuration are used to flexibly adjust parameters according to the scenario to maximize the efficiency of equipment use and lay the foundation for subsequent management; based on in-depth analysis of equipment operation data, relying on real-time monitoring and automatic configuration optimization, full life cycle management is achieved, the effective use time of equipment is extended, and a "monitoring-optimization" closed loop is formed; a three-dimensional digital twin model is introduced to intuitively display the equipment status, health status and configuration suggestions, and create an interactive interface. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is the system architecture framework diagram of the present invention, showing the architectural relationship between the patient demand analysis module, equipment status monitoring module, digital twin model construction module, and optimization decision and configuration module.

[0054] Figure 2 Flowchart of the algorithm for patient needs analysis, including the steps of keyword extraction, sentiment analysis, and generating patient needs reports.

[0055] Figure 3 This is a flow chart of the equipment condition monitoring algorithm, showing the process of data collection, transmission, analysis and anomaly detection.

[0056] Figure 4 Build a flowchart for the digital twin model to show the process of physical and behavioral attribute definition, real-time data fusion and visualization.

[0057] Figure 5 The flowchart of the device configuration optimization algorithm illustrates the logic of data collection, decision tree training, and configuration solution generation. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0059] Example 1

[0060] like Figure 1 As shown in the figure, the personalized medical device configuration and management system based on digital twin technology includes:

[0061] Patient demand analysis module: extracting patients’ individual needs and group common needs;

[0062] Equipment status monitoring module: collects equipment operation data in real time and uploads it to the cloud for analysis;

[0063] Digital twin model building module: creates a virtual model of the device, synchronizing the physical state and behavioral logic;

[0064] Optimization decision and configuration module: Generate optimal equipment configuration solutions based on data-driven.

[0065] In the present invention, personalized management and dynamic configuration are used to flexibly adjust parameters according to the scenario to maximize the efficiency of equipment use and lay the foundation for subsequent management; based on in-depth analysis of equipment operation data, relying on real-time monitoring and automatic configuration optimization, full life cycle management is achieved, the effective use time of equipment is extended, and a "monitoring-optimization" closed loop is formed; a three-dimensional digital twin model is introduced to intuitively display the equipment status, health status and configuration suggestions, and create an interactive interface.

[0066] Example 2

[0067] This embodiment provides a specific implementation method:

[0068] This invention uses digital twin technology to construct virtual models (digital twins) of medical devices, synchronizing data between the actual device and the digital model to support real-time status monitoring and data analysis. Digital twins, combined with IoT sensors and historical data analysis, can accurately reflect the current status of a device and predict future usage. This invention utilizes digital twin models to reflect the usage status of medical devices and patient needs in real time, optimizing device configuration through data analysis and intelligent decision-making to ensure the efficient use of medical resources.

[0069] System architecture framework: includes patient needs analysis module, equipment status monitoring module, digital twin model construction module, and optimization decision and configuration module.

[0070] Patient needs analysis module: used to collect basic information, medical history and treatment needs of patients. Analyze patient feedback and questionnaire data using natural language processing (NLP) technology: including the following steps (such as Figure 2 shown):

[0071] Keyword extraction: Use TF-IDF (Term Frequency-Inverse Document Frequency) or word vector method to extract high-frequency keywords in patient feedback. The TF-IDF formula is:

[0072] TF-IDF(t,d)=TF(t,d)×IDF(t)

[0073] Where TF(t,d) is the frequency of term t in document d, and IDF(t) is the inverse document frequency of term t.

[0074] Sentiment analysis: Clean and preprocess text data, such as removing stop words and punctuation, unifying text capitalization, and extracting stems; use deep learning models (such as LSTM, BERT) or traditional NLP methods (such as Naive Bayes classifier) to perform sentiment analysis on patient feedback.

[0075] Taking BERT as an example, the sentiment classification formula is:

[0076] Sentiment=argmax(BERT(f patient ))

[0077] Among them, f patient For patient feedback text, the BRET model outputs a probability distribution of sentiment labels; the patient's emotions are further quantified through sentiment scores (a numerical value ranging from -1 to 1).

[0078] Generate patient demand report: Based on sentiment analysis results and patient demand model, summarize the patient's specific needs into a demand report; use frequent item set mining (such as Apriori algorithm) to extract common needs, and the frequent item set support is calculated as:

[0079]

[0080] Where, the numerator (Number of transactions containing X) represents the number of transactions (such as patient demand records) containing demand item X; the denominator (Total number of transactions) represents the total number of transactions;

[0081] Formula meaning: Support (Support X) is the frequency of occurrence of requirement item X in all transactions, that is, the proportion of transactions containing X to the total number of transactions. For example, if there are 100 transactions in total, 30 of which contain X, then Indicates that X appears in 30% of transactions; by setting a minimum support threshold (such as 0.2), the set of demand items that appear frequently can be screened out (when 0.3>0.2, X meets the condition), providing a basis for the Apriori algorithm to mine common needs of patients.

[0082] Equipment status monitoring module: used to monitor the use status of medical equipment in real time through sensors, including equipment operating time, fault records, etc. (such as Figure 3 shown).

[0083] Transmit data to the cloud via IoT protocols (such as MQTT and HTTP) and use a time series database (such as InfluxDB) to store time series data.

[0084] Perform data analysis to identify device status: Use time series analysis or anomaly detection algorithms, such as moving average detection, using the following formula:

[0085]

[0086] Among them, MA t is the moving average at time t, x i For the i-th data point in the data sequence, a reasonable threshold range is set. If the sensor data exceeds the set threshold range, an alarm signal is triggered.

[0087] Digital twin model building modules: including defining the physical properties and behaviors of the device model, updating the model status in combination with real-time data, and using visualization tools to display the model (such as Figure 4 shown);

[0088] The physical properties and behaviors of the device model are defined as follows: Physical properties include basic properties such as device size, material, and operating temperature. Physical formulas are used to describe the performance parameters of the device. For example, the heat conduction equation (Fourier's law) is:

[0089]

[0090] Where q is the heat flux (W / m 2 ), which represents the heat transfer rate per unit area; k is the thermal conductivity (W / (m·K)); is the temperature gradient, describing the change of temperature with position. The behavioral properties define the response behavior of the device in different states. The temperature change model uses a differential equation to model the rate of change of the device temperature with time, expressed as:

[0091]

[0092] Where T(t) is the temperature of the device at time t; T ambient is the ambient temperature; α is the temperature transfer coefficient, which depends on the material and structure of the device. State change logic: if the device temperature exceeds a certain threshold Tmax, it will enter the "overheating" state, that is:

[0093] if T(t)>T max ,then state=overheat

[0094] When a device enters an "overheat" state, it can respond by automatically adjusting (such as reducing power) or issuing an alarm.

[0095] For example, consider a device whose temperature changes over time. The temperature change is affected by internal heat sources and the external environment. The temperature change rate can be expressed as:

[0096]

[0097] Where Q internal is the heat generated by the heat source inside the device (W); m is the mass of the device (kg); c is the specific heat capacity (J / kg·K); α is the cooling coefficient, which represents the heat transfer rate between the device and the environment.

[0098] Update the model state in combination with real-time data: Update the model state in combination with real-time data: Update the state variables of the digital twin model through sensor data, and use methods such as Kalman filter to predict and update real-time data. The formula is:

[0099]

[0100] in, is the state after the model is updated, K is the gain, z k is the measured value.

[0101] Use visualization tools to display models: Use 3D visualization tools (such as BowellTwin, Unity, Three.js, and WebGL) to display the real-time status of the device model and show the device's actions and status changes in a virtual environment.

[0102] Optimization decision and configuration module: used to collect data on device usage (frequency of use, usage time) and patient needs (device preferences, feedback) (such as Figure 5 shown).

[0103] Use decision tree algorithm for configuration decision: Feature selection in decision tree construction is based on information gain, which is calculated as:

[0104]

[0105] Where H(D) is the entropy of the data set D, D v The essence of information gain InfoGain(D,A) is: the uncertainty H(D) of the original data set D and the average uncertainty of each subset after being divided by feature A. The larger the difference, the greater the contribution of feature A to reducing data uncertainty, and the more suitable it is as a partitioning feature.

[0106] In a specific embodiment.

[0107] Output optimized equipment configuration plan: Based on information such as patient needs and equipment utilization, adjust equipment configuration to increase the number of high-demand equipment and reduce idle equipment; verify the optimized equipment configuration through simulation model to ensure the rationality of the optimization plan.

[0108] Patient information, including medical records, feedback, questionnaires, and physician notes, is obtained from the database. TF-IDF is used to extract high-frequency keywords, such as "pain relief" and "easy operation." After preprocessing the patient feedback, it is input into the BERT model for sentiment analysis, which outputs a probability distribution of sentiment labels. The sentiment category with the highest probability is selected and a sentiment score is generated. Frequent item sets are mined using the Apriori algorithm, with a minimum support threshold of 0.5. If the support for {"easy operation" and "real-time feedback"} is 0.6, it is determined to be a common need. Finally, a report is generated that includes both common needs and individual specific needs.

[0109] In a specific embodiment, the device status monitoring module is implemented as follows:

[0110] The sensor collects real-time temperature, humidity, and vibration data from the device and transmits it to the cloud-based InfluxDB database via the MQTT protocol. A moving average is calculated on the temperature data, with a window size of 10. If the temperature exceeds a threshold (e.g., above 80°C) at a given moment, an alarm is triggered, indicating that the device may be overheating.

[0111] In a specific embodiment, the digital twin model construction module is implemented as follows:

[0112] Define the physical properties of the device, including size and material, and use the heat conduction equation to describe temperature changes. Real-time temperature data is fed into the model, and state variables are updated using a Kalman filter. The device's temperature changes are displayed in real time in a Three.js visualization interface. When the temperature exceeds Tmax (70°C), the virtual model displays a red alert.

[0113] In a specific embodiment, the optimization decision and configuration module is implemented as follows:

[0114] Data on device usage frequency and patient needs were collected. Frequency of device use and patient need level were selected as features to train a decision tree model. Information gain was calculated to determine the optimal segmentation features, such as using a frequency of use ≥ 5 times / day as the cutoff point. Configuration rules were then generated. Device configurations were adjusted based on these rules, including increasing the number of high-demand devices. Simulation models were used to verify the rationality of the configuration plan and ensure efficient resource utilization.

[0115] In summary, the present invention realizes the personalized configuration and full life cycle management of medical equipment through the combination of digital twin technology, Internet of Things sensors, natural language processing, machine learning and other technologies, effectively improving the utilization efficiency and intelligent management level of medical resources.

[0116] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0117] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. Personalized medical equipment configuration and management system based on digital twin technology, characterized by: include: Patient demand analysis module: extracting patients’ individual needs and group common needs; Equipment status monitoring module: collects equipment operation data in real time and uploads it to the cloud for analysis; Digital twin model building module: creates a virtual model of the device, synchronizing the physical state and behavioral logic; Optimization decision and configuration module: Generate optimal equipment configuration solutions based on data-driven.

2. The personalized medical device configuration and management system based on digital twin technology according to claim 1, characterized in that: The patient needs analysis module includes the following aspects: Data acquisition: Extracting basic patient information from the database, including patient medical records, feedback, questionnaires, and doctor's notes, and using NLP to extract high-frequency keywords in patient feedback; Sentiment analysis: Use NLP methods to perform sentiment analysis on patient feedback to obtain sentiment labels and sentiment scores; Generate patient demand reports: Based on the previous sentiment analysis results, summarize the patient's specific needs into a demand report, identify high-frequency demand combinations through the Apriori algorithm, and generate individual and group demand reports.

3. The personalized medical device configuration and management system based on digital twin technology according to claim 2, characterized in that: The specific high-frequency keywords extracted from patient feedback using NLP are: in, is the frequency of term t in document d, is the inverse document frequency of term t.

4. The personalized medical device configuration and management system based on digital twin technology according to claim 2, characterized in that: Use NLP methods to perform sentiment analysis on patient feedback and obtain sentiment labels and sentiment scores as follows: Data preprocessing: Clean and preprocess text data to remove stop words and punctuation, unify text case, and extract stems; Sentiment Analysis Model: Use the BERT model to perform sentiment analysis on patient feedback. We assign sentiment labels, representing different sentiment directions by setting positive, neutral, and negative. The processed patient feedback is input into the BERT model, which then outputs the probability distribution of each sentiment label. The emotion category is selected based on the maximum probability, as shown in the following formula: in, For patient feedback text, the BRET model outputs a probability distribution of sentiment labels; the patient's emotions are further quantified through sentiment scores (a numerical value ranging from -1 to 1).

5. The personalized medical device configuration and management system based on digital twin technology according to claim 2, characterized in that: The Apriori algorithm identifies high-frequency demand combinations and generates individual and group demand reports as follows: Step S1, constructing a patient demand dataset; converting each patient's demand into a set of items and constructing a database, where each record represents a patient's demand set; In step S2, the Apriori algorithm is used to extract the common needs of the patient group. Specifically, the Apriori algorithm is used to traverse all patient need sets and calculate the support of each need item. The minimum support threshold is set to filter out the item sets that meet the conditions, as shown in the following formula: Where, the numerator (Number of transactions containing X) represents the number of transactions containing demand item X; the denominator (Total number of transactions) represents the total number of transactions; Step S3: Combine the frequent itemsets from the previous round to generate new candidate itemsets, calculate their support, and continue to screen itemsets whose support meets the threshold; Step S4, repeat the above process until no new frequent item sets can be found; the frequent item sets finally generated are the common needs of the patient group.

6. The personalized medical device configuration and management system based on digital twin technology according to claim 1, characterized in that: Equipment condition monitoring includes: Real-time data collection: Obtain equipment operating parameters through sensors, including equipment temperature, vibration, and usage time; Data transmission and storage: Upload data to the cloud time series database via IoT protocols (MQTT / HTTP); Anomaly detection: Use the moving average method to identify abnormal equipment status and trigger an early warning.

7. The personalized medical device configuration and management system based on digital twin technology according to claim 6, characterized in that: The moving average method is used to identify abnormal equipment status and trigger early warnings as follows: Where, is the moving average at time t, is the i-th data point in the data sequence. If the sensor data exceeds the set threshold range, an alarm signal is triggered.

8. The personalized medical device configuration and management system based on digital twin technology according to claim 1, characterized in that: Digital twin model construction includes: Physical and behavioral modeling: Define the physical properties and behavioral logic of the device. Physical properties include device size and material. Behavioral logic includes differential equations for degree change and state transition conditions. Real-time data fusion: Integrate sensor data through Kalman filter and dynamically update virtual model status; 3D visualization: Use 3D tools to display the real-time status of equipment and support interactive monitoring.

9. The personalized medical device configuration and management system based on digital twin technology according to claim 1, characterized in that: Optimization decisions and configurations include: Data input: Collect device usage and patient needs data; integrate device usage data with patient needs level; Intelligent decision-making: The configuration model is trained through the decision tree algorithm, and key features are selected based on information gain; Solution output: Generate dynamic configuration suggestions and verify their rationality through simulation.

10. The personalized medical device configuration and management system based on digital twin technology according to claim 9, characterized in that: The configuration model trained by the decision tree algorithm is as follows: First, we select the key features that affect device configuration. Then, we train a decision tree based on these key features to determine the configuration rules, as shown in the following formula: Where, is the entropy of the data set D, is a dataset with feature A value v; information gain The essence of is: the uncertainty of the original data set D and the average uncertainty of each subset after division by feature A The larger the difference, the greater the contribution of feature A to reducing data uncertainty, and the more suitable it is as a partitioning feature.

Images