Method and system for predicting malnutrition of hemodialysis patient based on AI technology

Abstract

The invention provides an AI technology-based hemodialysis patient malnutrition prediction method and system. The system comprises a data collection module, a data preprocessing module, a data standardization module, a dimension reduction module and a nutrition state prediction module. The method comprises the following steps: S1, collecting related attributes of hemodialysis patients to obtain a hemodialysis patient data set; s2, preprocessing the hemodialysis patient data set; s3, performing standardization processing on the hemodialysis patient data set; s4, a feature selection and model construction method; s5, model training and generation; according to the method, the AI technology is combined with the multi-dimensional data to comprehensively analyze the patient, the nutrition state of the patient can be more comprehensively predicted, the evaluation result is comprehensive, and the problems that in the prior art, the nutrition state of the hemodialysis patient is predicted based on a few indexes, the indexes are too single, and the prediction result is not comprehensive are solved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for predicting malnutrition in hemodialysis patients based on AI technology. Background Art

[0002] Hemodialysis is a common method for treating chronic kidney disease, but long-term hemodialysis treatment may lead to malnutrition in patients, which in turn affects their quality of life and prognosis; therefore, it is very necessary to monitor and predict the nutritional status of hemodialysis patients; The existing technology usually predicts the nutritional status of hemodialysis patients based on a few indicators. The indicators are too single and cannot fully predict the nutritional status of patients, and the evaluation results lack comprehensiveness. At the same time, some indicators may be easily affected by other factors, resulting in inaccurate predictions of patients' nutritional status. Summary of the invention

[0003] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method and system for predicting malnutrition in hemodialysis patients based on AI technology.

[0004] To achieve the above object, the present invention adopts the following technical solutions: A system for predicting malnutrition in hemodialysis patients based on AI technology, including a data collection module, a data preprocessing module, a data standardization module, a dimension reduction module, and a nutritional status prediction module; The data collection module is used to collect data related to hemodialysis patients; The data preprocessing module is used to preprocess the data in the hemodialysis patient data set; The data standardization module is used to standardize the data in the hemodialysis patient data set; The dimension reduction module is used to perform cluster analysis on the data in the hemodialysis patient data set, remove features that are less correlated with malnutrition, and perform iterative clustering and grouping to obtain features that are highly correlated with malnutrition; The nutritional status prediction module is used to establish a prediction model and predict whether the patient is in a malnutrition state.

[0005] A method for predicting malnutrition in hemodialysis patients based on AI technology, comprising the following steps: S1: Collect the relevant attributes of hemodialysis patients to obtain the hemodialysis patient dataset; The data collection module collects the relevant attributes of hemodialysis patients according to actual needs; Specifically, the data collection module collects hemodialysis patient-related attributes from general demographic data, hemodialysis data, laboratory test indicators, and diagnostic results according to actual needs; The general demographic data include gender, age, body mass index (BMI), ethnicity, marital status, education level, whether there is hypertension, whether there is diabetes, primary disease type, post-dialysis systolic blood pressure, post-dialysis diastolic blood pressure, whether there is hematoma after dialysis, etc.; The hemodialysis data include hemodialysis time, 24-hour urine volume, blood flow, ultrafiltration volume, vascular access type, whether the access is infected, etc.; the vascular access type includes arteriovenous fistula, arteriovenous valve, central venous catheter, peritoneal dialysis catheter, etc.; The laboratory examination indicators include hemoglobin, total protein, prealbumin, blood sugar 2 hours after meal, blood creatinine, urea nitrogen, blood uric acid, etc.; The diagnosis result of the hemodialysis patient is a specific disease; The collection of hemodialysis patient-related attributes is completed to obtain a hemodialysis patient data set, which includes features and corresponding feature values; the features are various attributes of the collected patients, and the feature values ​​are specific values ​​corresponding to various attributes.

[0006] S2: Preprocessing of hemodialysis patient dataset; The data preprocessing module preprocesses the data in the hemodialysis patient dataset by processing missing data and outliers, including the following sub-steps: S21: Handling missing data; Determine whether there is missing data in the hemodialysis patient data set, and if so, process the missing data; if not, proceed to step S22; S211: Determine whether there is missing data; Determine whether there is missing data in the hemodialysis patient dataset; Specific judgment methods include statistical missing values, visual data distribution, etc. Count missing values: Use the isnull or isna function to identify missing data in a dataset; Visualize data distribution: Draw charts to show the distribution of data in a dataset and identify missing data; After the judgment is completed, the missing data is summarized through the sum function to obtain the total number of missing data in the data set; After the judgment is completed, the number of missing data is obtained through the sum function; If there is no missing data in the hemodialysis patient data set, proceed to step S22; if there is missing data, proceed to step S212; S212: Evaluate the impact of missing data; Assess the impact of missing data on the overall data distribution. Specific assessment methods include comparing statistical indicators, calculating correlations, and comparing data distributions; Specifically, comparing statistical indicators refers to calculating statistical indicators of a hemodialysis patient data set containing features corresponding to the missing data and statistical indicators of a hemodialysis patient data set without features corresponding to the missing data according to the missing data detected in step S211, wherein the statistical indicators include mean, median, standard deviation, etc.; comparing the relative difference or percentage difference of the calculated statistical indicators, if the difference is small, it means that the missing data has little impact on the overall data distribution; otherwise, it means that the missing data has a greater impact on the overall data distribution; Calculating the correlation means calculating the Pearson correlation coefficient between each feature in the hemodialysis patient data set by using the corr function to obtain a correlation matrix, and extracting the correlation coefficient between the feature corresponding to the missing data and other features in the correlation matrix. If the correlation coefficient is small, it means that the missing data has little impact on the overall data distribution; otherwise, it means that the missing data has a greater impact on the overall data distribution; Comparing data distribution means plotting the data distribution of the hemodialysis patient data set containing the features corresponding to the missing data and the data distribution of the hemodialysis patient data set without the features corresponding to the missing data through a histogram, and comparing the data distribution in the above two cases through a box plot. If the data distributions of the two cases are similar, the missing data has a small impact on the overall data distribution. Otherwise, the missing data has a large impact on the overall data distribution. S213: Processing of missing data; Process the missing data according to the evaluation result of step S212. When the amount of missing data is small and the missing data has a small impact on the overall data distribution, use the filling method or the deletion method; Furthermore, the filling method includes mean filling, median filling, mode filling, etc., and a suitable filling method is selected according to the data type and data distribution characteristics; The deletion method refers to deleting rows or columns containing missing data in the hemodialysis patient data set; When the amount of missing data is large and the missing data has a significant impact on the overall data distribution, deep learning methods, multiple interpolation methods, etc. are used to fill in the data; When there are missing data in continuous data, interpolation methods are used to fill in the missing data, such as linear interpolation or polynomial interpolation. S22: Handling outliers; Determine whether there are abnormal values ​​in the hemodialysis patient data set. If there are abnormal values, process the abnormal values; if not, proceed to step S23; S221: Identify outliers; Identify whether there are outliers in the hemodialysis patient data set. Specific identification methods include box plot method, Z-score comparison detection method, etc. Specifically, the box plot method refers to drawing a box plot, wherein the box plot contains the median, upper quartile, lower quartile, upper limit, and lower limit of the data in the hemodialysis patient data set; if there is a data point in the box plot that is higher than the upper limit or lower than the lower limit, the data corresponding to the data point is an outlier, otherwise it is a non-outlier; The Z-score comparison detection method refers to calculating the mean and standard deviation of the data in the hemodialysis patient data set, and calculating the standard score Z-score of each data point according to the calculated mean, standard deviation and the data value in the hemodialysis patient data set through the Z-score standardization formula, and setting a threshold in advance. If the standard score Z-score of the data point is greater than the threshold, it is an abnormal point, otherwise it is a non-abnormal point; the Z-score standardization formula is as follows: ; Among them, Z-score is the standard score, x is the data value in the hemodialysis patient data set, is the mean value and s is the standard deviation; S222: Processing abnormal values; Dealing with outliers by deleting or replacing them; Specifically, replacing outliers means replacing outliers with specific values ​​in the hemodialysis patient dataset, such as the median, mean, etc.; S3: Standardization of hemodialysis patient dataset; The data standardization module standardizes the data in the hemodialysis patient data set through data standardization; The diagnostic results of hemodialysis patients were standardized into mild malnutrition state, moderate malnutrition state, severe malnutrition state, and normal nutrition state; Furthermore, the hemodialysis patient dataset contains features of two types of tests: qualitative tests and quantitative tests; For the characteristics of qualitative detection, the detection result, i.e., the characteristic value, is represented by a numerical value, with a positive value of 1 and a negative value of 0; For the features of quantitative detection, the normal range of the detection results of this feature is determined in advance, the actual detection results are compared with the normal range, and the difference value of each feature is calculated; The normal range of the test result includes a single value and an interval; If the normal range of the test result is a single value, the difference value = |actual test result - normal range of the test result|; If the test result is an interval, calculate the center value of the interval, which is the average of the upper and lower limits of the interval, and the difference value = |actual test result-interval center value|; The mean and standard deviation of the difference values ​​are obtained based on the calculated difference values ​​of the features, and the data is standardized using the Z-score standardization formula; At this time, in the Z-score standardization formula, Z-score is the standard score, x is the difference value, is the mean of the difference values, and s is the standard deviation of the feature difference values.

[0007] S4: Feature selection and model building methods; The dimensionality reduction module combines data algorithms to identify features associated with malnutrition in the hemodialysis patient dataset; S41: Preliminary removal of features with small differences; Cluster analysis is performed on the data in the hemodialysis patient dataset using a clustering algorithm; The clustering algorithms include k-means clustering, hierarchical clustering, and DBSCAN clustering; The data of each patient in the hemodialysis patient data set is recorded as a sample, and the above three clustering algorithms are run on each sample respectively. Each clustering algorithm assigns each sample to the corresponding cluster respectively; The method for setting the number of clusters for each clustering algorithm includes: the number of clusters for k-means clustering and hierarchical clustering uses the silhouette coefficient to determine the optimal number of clusters, and DBSCAN clustering does not require the number of clusters to be set in advance, but does require the eps and min_samples parameters to be set, which are determined by observing the data point distance graph; Calculate the distribution of each feature in each cluster, and measure the distribution difference of each feature in different clusters by variance; Calculate the variance of each feature contained in each cluster according to the variance formula, arrange the calculated variance of each feature in descending order, and draw a curve after the variance is sorted, in which the horizontal axis of the curve is the feature name and the vertical axis is the variance of the corresponding feature; The kneele algorithm is used to obtain the slope change of the curve, and the inflection point of the curve is detected according to the slope change. The features before the first inflection point of the detected curve show greater differences between different clusters, and are more able to distinguish different clusters; The features that fail to show significant differences in different clusters in the data set, that is, the features after the first inflection point, are removed. The intersection of the features with significant differences obtained by the three clustering algorithms is the features after preliminary dimensionality reduction. The entire experimental process is iterative. The conditions for the end of the iteration can be determined based on the following three points: (1) The curve of feature variance ranking is stable: When the position of the first inflection point of the variance ranking curve calculated by the Kneedle algorithm no longer changes significantly in consecutive iterations, it indicates that the features that can distinguish different clusters have stabilized and the iteration is stopped; (2) Convergence of feature intersection of the three clustering algorithms: Compare the large difference feature sets obtained by K-means clustering, hierarchical clustering and DBSCAN. If the feature intersection gradually decreases after each iteration and the feature set of the final intersection no longer changes significantly, it means that a stable preliminary dimensionality reduction feature has been found and the iteration can be stopped. (3) Clustering performance indicators meet expectations: Evaluate the clustering effect after each iteration, using indicators such as silhouette coefficient, purity of nutrient status within the cluster, and accuracy. If these indicators do not improve significantly over multiple iterations, it indicates that the current feature set is effective enough and no further feature removal is required. If at least two of the above conditions are met, it can be considered that the iterative process of S41 has reached the state of optimal feature screening; S42: Data clustering calculation generation model; S41 initially removes features with small differences to obtain new features. Each cluster obtained by each clustering algorithm in S41 retains the new features. For each sample, a vote is conducted based on the clustering results of K-means, hierarchical clustering, and DBSCAN, and the most frequent cluster is selected as the final result. The cluster labels of the three algorithms are used as new features and added to the data set to further improve the clustering effect. Label each patient in each cluster with a nutritional status; the nutritional status includes normal nutrition and malnutrition; the malnutrition status label is set to 1, and the malnutrition status includes mild, moderate, and severe malnutrition; the normal nutrition status label is set to 0; For patients in each cluster, statistics were taken based on their nutritional status, and the proportion of people in each nutritional status in the cluster was calculated; Label each cluster with a nutritional status label, which is the nutritional status with the highest proportion in each cluster; this step is to define a representative nutritional status for each cluster. These labels can help understand the differences between clusters; Finally, a clustering model was generated. The key indicators for evaluating the clustering model included the silhouette coefficient, the purity and accuracy of the nutritional status within the cluster. If the silhouette coefficient was above 0.5, the purity of the nutritional status within the cluster was above 80%, and the accuracy was above 85%, it indicated that the clustering model could be used to identify the nutritional status of new data. S43: Randomly combine features for iterative clustering grouping; S43 is an optimization of step S42. In step S42, a representative nutritional status label has been assigned to each cluster, and the nutritional status ratio of each cluster has been calculated. In order to further optimize the clustering model and feature selection, S43 performs multiple iterations of clustering analysis by randomly combining features. From the features initially screened in step S41, feature subsets of different combinations are randomly selected and used to re-cluster the data; in each clustering, k-means clustering, hierarchical clustering and DBSCAN clustering methods are reapplied, the previous cluster number determination and clustering steps are repeated, and the results are fused; in each iteration, the process consistent with S42 is maintained, that is, a nutritional status label is assigned to each cluster, and the nutritional status ratio in each cluster is calculated; Iterative clustering is performed by randomly combining different features multiple times, with the aim of observing the impact of different feature combinations on clustering results and finding the feature combination that can best distinguish nutritional status; After each clustering, the effect of the clustering model was evaluated to ensure that the model could effectively distinguish different nutritional states. The silhouette coefficient of each cluster, the purity and accuracy of the nutritional state within the cluster were calculated to ensure that the clustering results could effectively distinguish the nutritional states. When similar feature combinations in multiple iterations repeatedly obtain similar clustering results, and the indicators such as the purity of the nutrient status within the cluster, the silhouette coefficient and the accuracy rate reach the set threshold, that is, when the silhouette coefficient remains stable above 0.5, the purity value remains above 80%, and the accuracy rate remains above 85%, and there is no significant improvement in further iterations, the iteration can be stopped; Through S43 iterative optimization, the feature set can be further narrowed down, and the most discriminative feature combination can be finally selected to improve the recognition accuracy of the clustering model for new data; Characteristics found to be highly associated with malnutrition in existing studies remained constant throughout the analysis; The dimensionality reduction method proposed in this invention starts from the data structure and captures the relationship between features and specific subgroups (clusters), which helps to retain the key features in distinguishing these subgroups; traditional dimensionality reduction methods usually reduce dimensions by transforming data or based on linear algebra methods without involving clustering processes; S5: Model training and generation; The nutritional status prediction module is based on the feature selection and cluster analysis of S4. It predicts the malnutrition status through the final screened features and clustering model, including the following sub-steps: S51: prepare training data; The feature selection results of step S4 are used to determine the feature set. These features are the most effective features for distinguishing malnutrition status after multiple clustering iterations. Each sample has been labeled with a nutritional status label (0 indicates normal nutrition, 1 indicates malnutrition), that is, mild, moderate, and severe malnutrition are classified as 1, and normal nutrition is 0; S52: Use clustering models to predict whether a patient is in a malnourished state; The patient samples have been assigned to different clusters through the previous clustering models (S42, S43), and a nutritional status label has been assigned to each cluster. This clustering structure is used to build a prediction model.

[0008] S53: Training using clustering model; The filtered feature set (the final feature combination in step S43) is used as input data.

[0009] Based on the characteristics of each new sample, the trained clustering model is used to assign the new sample to a cluster. Using the nutritional status ratio previously calculated for each cluster, the new sample is given a predicted nutritional status label. Specifically, the new sample will be assigned to a cluster, and the nutritional status ratio in the cluster determines the label of the sample (0 or 1). The prediction accuracy is evaluated using the known nutritional status labels (0 or 1 marked in the training set). The performance of the model can be evaluated by accuracy, precision, recall, F1 score and other criteria; S54: Model evaluation and optimization; Evaluation metrics include accuracy, precision, recall, and F1 score. If the initial training results are not satisfactory, adjust the parameters of the clustering algorithm (such as eps and min_samples in DBSCAN, or the number of clusters in K-means), or re-combine features during feature selection. S55: Use the trained clustering model to predict new data; For each new patient, after inputting their characteristics, the clustering model assigns the patient to a cluster and predicts whether the patient is in a malnutrition state based on the nutritional status label of the cluster. Through the prediction results of the model, malnourished patients can be intervened in advance and corresponding nutritional support or other treatments can be provided.

[0010] Compared with the prior art, the present invention has the following beneficial effects: The method of the present invention uses artificial intelligence (AI) technology combined with multidimensional data to conduct a comprehensive analysis of patients, which can more comprehensively predict the nutritional status of patients. The evaluation results are comprehensive, avoiding the problems of predicting the nutritional status of hemodialysis patients based on a few indicators in the prior art, with too few indicators and incomplete prediction results. The method of the present invention avoids the problem of limitations of a single algorithm, and improves the accuracy of obtaining factors affecting malnutrition by compensating each other through multiple algorithms, comparing in multiple ways, and verifying each other; the results of multiple algorithms may have differences or deviations, and by cross-validating and comparing the results of multiple algorithms, they can compensate each other and ensure the accuracy of obtaining factors affecting malnutrition. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a flowchart of the steps of a method for predicting malnutrition in hemodialysis patients based on AI technology of the present invention; Figure 2 This is a module function implementation diagram of a method for predicting malnutrition in hemodialysis patients based on AI technology in the present invention. DETAILED DESCRIPTION

[0012] In order to provide a further understanding of the purpose, structure, features, and functions of the present invention, the following detailed description is given in conjunction with the embodiments.

[0013] like Figure 1 and Figure 2 As shown, a system for predicting malnutrition in hemodialysis patients based on AI technology includes a data collection module, a data preprocessing module, a data standardization module, a dimension reduction module, and a nutritional status prediction module; The data collection module is used to collect data related to hemodialysis patients; The data preprocessing module is used to preprocess the data in the hemodialysis patient data set; The data standardization module is used to standardize the data in the hemodialysis patient data set; The dimension reduction module is used to perform cluster analysis on the data in the hemodialysis patient data set, remove features that are less correlated with malnutrition, and perform iterative clustering and grouping to obtain features that are highly correlated with malnutrition; The nutritional status prediction module is used to establish a prediction model and predict whether the patient is in a malnutrition state.

[0014] A method for predicting malnutrition in hemodialysis patients based on AI technology, comprising the following steps: S1: Collect the relevant attributes of hemodialysis patients to obtain the hemodialysis patient dataset; The data collection module collects the relevant attributes of hemodialysis patients according to actual needs; Specifically, the data collection module collects hemodialysis patient-related attributes from general demographic data, hemodialysis data, laboratory test indicators, and diagnostic results according to actual needs; The general demographic data include gender, age, body mass index (BMI), ethnicity, marital status, education level, whether there is hypertension, whether there is diabetes, primary disease type, post-dialysis systolic blood pressure, post-dialysis diastolic blood pressure, whether there is hematoma after dialysis, etc.; The hemodialysis data include hemodialysis time, 24-hour urine volume, blood flow, ultrafiltration volume, vascular access type, whether the access is infected, etc.; the vascular access type includes arteriovenous fistula, arteriovenous valve, central venous catheter, peritoneal dialysis catheter, etc.; The laboratory examination indicators include hemoglobin, total protein, prealbumin, blood sugar 2 hours after meal, blood creatinine, urea nitrogen, blood uric acid, etc.; The diagnosis result of the hemodialysis patient is a specific disease; After the collection of the hemodialysis patient-related attributes is completed, a hemodialysis patient data set is obtained, wherein the hemodialysis patient data set includes features and corresponding feature values; the features are various attributes of the collected patients, and the feature values ​​are specific values ​​corresponding to various attributes; Furthermore, the inclusion and exclusion criteria were determined; The inclusion and exclusion criteria include inclusion criteria and exclusion criteria; The inclusion criteria refer to the conditions that must be met for a patient to be included in the study or treatment, including the following: (1) Aged ≥ 18 years; (2) diagnosed with end-stage renal disease; (3) The patient has been on regular dialysis (2-4 times / week) for ≥3 months in this blood purification center; (4) Complete dialysis data for at least 1 month; The exclusion criteria refer to the conditions that should be excluded from patient research or treatment, including the following: (1) Patients with cardiovascular and cerebrovascular diseases, infections, malignant tumors, gastrointestinal and liver diseases, tuberculosis, arrhythmias, etc.; (2) Incomplete clinical and experimental data; According to the inclusion and exclusion criteria, the patient data that met the requirements in the hemodialysis patient data set were screened; By incorporating exclusion criteria to screen out patient data that meet the requirements, the accuracy and credibility of the data can be improved, unnecessary data processing and analysis can be reduced, and the efficiency of the research can be improved.

[0015] S2: Preprocessing of hemodialysis patient dataset; The data preprocessing module preprocesses the data in the hemodialysis patient dataset by processing missing data, processing outliers, and feature selection, including the following sub-steps: S21: Handling missing data; Determine whether there is missing data in the hemodialysis patient data set, and if so, process the missing data; if not, proceed to step S22; S211: Determine whether there is missing data; Determine whether there is missing data in the hemodialysis patient dataset; Specific judgment methods include statistical missing values, visual data distribution, etc. Count missing values: Use the isnull or isna function to identify missing data in a dataset; Visualize data distribution: Draw charts to show the distribution of data in a dataset and identify missing data; After the judgment is completed, the missing data is summarized through the sum function to obtain the total number of missing data in the data set; After the judgment is completed, the number of missing data is obtained through the sum function; If there is no missing data in the hemodialysis patient data set, proceed to step S22; if there is missing data, proceed to step S212; S212: Evaluate the impact of missing data; Assess the impact of missing data on the overall data distribution. Specific assessment methods include comparing statistical indicators, calculating correlations, and comparing data distributions; Specifically, comparing statistical indicators refers to calculating statistical indicators of a hemodialysis patient data set containing features corresponding to the missing data and statistical indicators of a hemodialysis patient data set without features corresponding to the missing data according to the missing data detected in step S211, wherein the statistical indicators include mean, median, standard deviation, etc.; comparing the relative difference or percentage difference of the calculated statistical indicators, if the difference is small, it means that the missing data has little impact on the overall data distribution; otherwise, it means that the missing data has a greater impact on the overall data distribution; Calculating the correlation means calculating the Pearson correlation coefficient between each feature in the hemodialysis patient data set by using the corr function to obtain a correlation matrix, and extracting the correlation coefficient between the feature corresponding to the missing data and other features in the correlation matrix. If the correlation coefficient is small, it means that the missing data has little impact on the overall data distribution; otherwise, it means that the missing data has a greater impact on the overall data distribution; Comparing data distribution means plotting the data distribution of the hemodialysis patient data set containing the features corresponding to the missing data and the data distribution of the hemodialysis patient data set without the features corresponding to the missing data through a histogram, and comparing the data distribution in the above two cases through a box plot. If the data distributions of the two cases are similar, the missing data has a small impact on the overall data distribution. Otherwise, the missing data has a large impact on the overall data distribution. S213: Processing of missing data; When the number of missing data is small and the missing data has a small impact on the overall data distribution, for example, when the missing data accounts for less than 10% of the total data in the hemodialysis patient dataset, the filling method or deletion method is used; Furthermore, the filling method includes mean filling, median filling, mode filling, etc., and a suitable filling method is selected according to the data type and data distribution characteristics; The deletion method refers to deleting rows or columns containing missing data in the hemodialysis patient data set; When the amount of missing data is large and the missing data has a significant impact on the overall data distribution, deep learning methods, multiple interpolation methods, etc. are used to fill in the data; When there are missing data in continuous data, interpolation methods are used to fill in the missing data, such as linear interpolation or polynomial interpolation. By interpolating and filling missing data, the continuity of data can be ensured, which is convenient for subsequent data analysis and other applications; S22: Handling outliers; Determine whether there are abnormal values ​​in the hemodialysis patient data set. If there are abnormal values, process the abnormal values; if not, proceed to step S23; S221: Identify outliers; Identify whether there are outliers in the hemodialysis patient data set. Specific identification methods include box plot method, Z-score comparison detection method, etc. Specifically, the box plot method refers to drawing a box plot, wherein the box plot contains the median, upper quartile, lower quartile, upper limit, and lower limit of the data in the hemodialysis patient data set; if there is a data point in the box plot that is higher than the upper limit or lower than the lower limit, the data corresponding to the data point is an outlier, otherwise it is a non-outlier; The Z-score comparison detection method refers to calculating the mean and standard deviation of the data in the hemodialysis patient data set, and calculating the standard score Z-score of each data point according to the calculated mean, standard deviation and the data value in the hemodialysis patient data set through the Z-score standardization formula, and setting a threshold in advance. If the standard score Z-score of the data point is greater than the threshold, it is an abnormal point, otherwise it is a non-abnormal point; the Z-score standardization formula is as follows: ; Among them, Z-score is the standard score, x is the data value in the hemodialysis patient data set, is the mean value and s is the standard deviation; S222: Processing abnormal values; Dealing with outliers by deleting or replacing them; Specifically, replacing outliers means replacing outliers with specific values ​​in the hemodialysis patient dataset, such as the median, mean, etc.; Furthermore, outliers can be handled by trimming the mean, median absolute deviation, and other methods; By dealing with outliers by trimming the mean, median absolute deviation and other methods, we can better resist the influence of outliers and make the description of the overall characteristics of the data more accurate; S3: Standardization of hemodialysis patient dataset; The data standardization module standardizes the data in the hemodialysis patient data set through data standardization; The diagnostic results of hemodialysis patients were standardized into mild malnutrition state, moderate malnutrition state, severe malnutrition state, and normal nutrition state; Furthermore, the hemodialysis patient dataset contains features of two types of tests: qualitative tests and quantitative tests; For the characteristics of qualitative detection, the detection result, i.e., the characteristic value, is represented by a numerical value, with a positive value of 1 and a negative value of 0; For the features of quantitative detection, the normal range of the detection results of this feature is determined in advance, the actual detection results are compared with the normal range, and the difference value of each feature is calculated; The normal range of the test result includes a single value and an interval; If the normal range of the test result is a single value, the difference value = |actual test result - normal range of the test result|; If the test result is an interval, calculate the center value of the interval, which is the average of the upper and lower limits of the interval, and the difference value = |actual test result-interval center value|; The mean and standard deviation of the difference values ​​are obtained based on the calculated difference values ​​of the features, and the data is standardized using the Z-score standardization formula; At this time, in the Z-score standardization formula, Z-score is the standard score, x is the difference value, is the mean of the difference values, and s is the standard deviation of the feature difference values.

[0016] S4: Feature selection and model building methods; The dimensionality reduction module combines data algorithms to identify features associated with malnutrition in the hemodialysis patient dataset; S41: Preliminary removal of features with small differences; Cluster analysis is performed on the data in the hemodialysis patient dataset using a clustering algorithm; The clustering algorithms include k-means clustering, hierarchical clustering, and DBSCAN clustering; The data of each patient in the hemodialysis patient data set is recorded as a sample, and the above three clustering algorithms are run on each sample respectively. Each clustering algorithm assigns each sample to the corresponding cluster respectively; The method for setting the number of clusters for each clustering algorithm includes: the number of clusters for k-means clustering and hierarchical clustering uses the silhouette coefficient to determine the optimal number of clusters, and DBSCAN clustering does not require the number of clusters to be set in advance, but does require the eps and min_samples parameters to be set, which are determined by observing the data point distance graph; Calculate the distribution of each feature in each cluster, and measure the distribution difference of each feature in different clusters by variance; Calculate the variance of each feature contained in each cluster according to the variance formula, arrange the calculated variance of each feature in descending order, and draw a curve after the variance is sorted, in which the horizontal axis of the curve is the feature name and the vertical axis is the variance of the corresponding feature; The kneele algorithm is used to obtain the slope change of the curve, and the inflection point of the curve is detected according to the slope change. The features before the first inflection point of the detected curve show greater differences between different clusters, and are more able to distinguish different clusters; The features that fail to show significant differences in different clusters in the data set, that is, the features after the first inflection point, are removed. The intersection of the features with significant differences obtained by the three clustering algorithms is the features after preliminary dimensionality reduction. The entire experimental process is iterative. The conditions for the end of the iteration can be determined based on the following three points: (1) The curve of feature variance ranking is stable: When the position of the first inflection point of the variance ranking curve calculated by the Kneedle algorithm no longer changes significantly in consecutive iterations, it indicates that the features that can distinguish different clusters have stabilized and the iteration is stopped; (2) Convergence of feature intersection of the three clustering algorithms: Compare the large difference feature sets obtained by K-means clustering, hierarchical clustering and DBSCAN. If the feature intersection gradually decreases after each iteration and the feature set of the final intersection no longer changes significantly, it means that a stable preliminary dimensionality reduction feature has been found and the iteration can be stopped. (3) Clustering performance indicators meet expectations: Evaluate the clustering effect after each iteration, using indicators such as silhouette coefficient, purity of nutrient status within the cluster, and accuracy. If these indicators do not improve significantly over multiple iterations, it indicates that the current feature set is effective enough and no further feature removal is required. If at least two of the above conditions are met, it can be considered that the iterative process of S41 has reached the state of optimal feature screening; S42: Data clustering calculation generation model; S41 initially removes features with small differences to obtain new features. Each cluster obtained by each clustering algorithm in S41 retains the new features. For each sample, a vote is conducted based on the clustering results of K-means, hierarchical clustering, and DBSCAN, and the most frequent cluster is selected as the final result. The cluster labels of the three algorithms are used as new features and added to the data set to further improve the clustering effect. Label each patient in each cluster with a nutritional status; the nutritional status includes normal nutrition and malnutrition; the malnutrition status label is set to 1, and the malnutrition status includes mild, moderate, and severe malnutrition; the normal nutrition status label is set to 0; For patients in each cluster, statistics were taken based on their nutritional status, and the proportion of people in each nutritional status in the cluster was calculated; Label each cluster with a nutritional status label, which is the nutritional status with the highest proportion in each cluster; this step is to define a representative nutritional status for each cluster. These labels can help understand the differences between clusters; Finally, a clustering model was generated. The key indicators for evaluating the clustering model included the silhouette coefficient, the purity and accuracy of the nutritional status within the cluster. If the silhouette coefficient was above 0.5, the purity of the nutritional status within the cluster was above 80%, and the accuracy was above 85%, it indicated that the clustering model could be used to identify the nutritional status of new data. S43: Randomly combine features for iterative clustering grouping; S43 is an optimization of step S42. In step S42, a representative nutritional status label has been assigned to each cluster, and the nutritional status ratio of each cluster has been calculated. In order to further optimize the clustering model and feature selection, S43 performs multiple iterations of clustering analysis by randomly combining features. From the features initially screened in step S41, feature subsets of different combinations are randomly selected and used to re-cluster the data; in each clustering, k-means clustering, hierarchical clustering and DBSCAN clustering methods are reapplied, the previous cluster number determination and clustering steps are repeated, and the results are fused; in each iteration, the process consistent with S42 is maintained, that is, a nutritional status label is assigned to each cluster, and the nutritional status ratio in each cluster is calculated; Iterative clustering is performed by randomly combining different features multiple times, with the aim of observing the impact of different feature combinations on clustering results and finding the feature combination that can best distinguish nutritional status; After each clustering, the effect of the clustering model was evaluated to ensure that the model could effectively distinguish different nutritional states. The silhouette coefficient of each cluster, the purity and accuracy of the nutritional state within the cluster were calculated to ensure that the clustering results could effectively distinguish the nutritional states. When similar feature combinations in multiple iterations repeatedly obtain similar clustering results, and the indicators such as the purity of the nutrient status within the cluster, the silhouette coefficient and the accuracy rate reach the set threshold, that is, when the silhouette coefficient remains stable above 0.5, the purity value remains above 80%, and the accuracy rate remains above 85%, and there is no significant improvement in further iterations, the iteration can be stopped; Through S43 iterative optimization, the feature set can be further narrowed down, and the most discriminative feature combination can be finally selected to improve the recognition accuracy of the clustering model for new data; Characteristics found to be highly associated with malnutrition in existing studies remained constant throughout the analysis; The dimensionality reduction method proposed in this invention starts from the data structure and captures the relationship between features and specific subgroups (clusters), which helps to retain the key features in distinguishing these subgroups; traditional dimensionality reduction methods usually reduce dimensions by transforming data or based on linear algebra methods without involving clustering processes; S5: Model training and generation; The nutritional status prediction module is based on the feature selection and cluster analysis of S4. It predicts the malnutrition status through the final screened features and clustering model, including the following sub-steps: S51: prepare training data; The feature selection results of step S4 are used to determine the feature set. These features are the most effective features for distinguishing malnutrition status after multiple clustering iterations. Each sample has been labeled with a nutritional status label (0 indicates normal nutrition, 1 indicates malnutrition), that is, mild, moderate, and severe malnutrition are classified as 1, and normal nutrition is 0; S52: Use clustering models to predict whether a patient is in a malnourished state; The patient samples have been assigned to different clusters through the previous clustering models (S42, S43), and a nutritional status label has been assigned to each cluster. This clustering structure is used to build a prediction model.

[0017] S53: Training using clustering model; The filtered feature set (the final feature combination in step S43) is used as input data.

[0018] Based on the characteristics of each new sample, the trained clustering model is used to assign the new sample to a cluster. Using the nutritional status ratio previously calculated for each cluster, the new sample is given a predicted nutritional status label. Specifically, the new sample will be assigned to a cluster, and the nutritional status ratio in the cluster determines the label of the sample (0 or 1). The prediction accuracy is evaluated using the known nutritional status labels (0 or 1 marked in the training set). The performance of the model can be evaluated by accuracy, precision, recall, F1 score and other criteria; S54: Model evaluation and optimization; Evaluation metrics include accuracy, precision, recall, and F1 score. If the initial training results are not satisfactory, adjust the parameters of the clustering algorithm (such as eps and min_samples in DBSCAN, or the number of clusters in K-means), or re-combine features during feature selection. S55: Use the trained clustering model to predict new data; For each new patient, after inputting their characteristics, the clustering model assigns the patient to a cluster and predicts whether the patient is in a malnutrition state based on the nutritional status label of the cluster. Through the prediction results of the model, malnourished patients can be intervened in advance and corresponding nutritional support or other treatments can be provided.

[0019] The method of the present invention preprocesses the data in the hemodialysis patient data set by processing missing data, processing outliers, and selecting features, and determines the influence of the corresponding data on the overall data distribution by calculating the Pearson correlation coefficient between each feature in the data set, thereby avoiding the problem that some indicators may be susceptible to the influence of other factors, resulting in inaccurate prediction of the patient's nutritional status.

[0020] The present invention has been described by the above-mentioned relevant embodiments, however, the above-mentioned embodiments are only examples for implementing the present invention. It must be pointed out that the disclosed embodiments do not limit the scope of the present invention. On the contrary, changes and modifications made without departing from the spirit and scope of the present invention are all within the scope of patent protection of the present invention.

Claims

1. A method for predicting malnutrition in hemodialysis patients based on AI technology, characterized by: The following steps are involved: S1: Collect the relevant attributes of hemodialysis patients to obtain the hemodialysis patient dataset; The data collection module collects the relevant attributes of hemodialysis patients according to actual needs; Specifically, the data collection module collects hemodialysis patient-related attributes from general demographic data, hemodialysis data, laboratory test indicators, and diagnostic results according to actual needs; The hemodialysis patient related attributes are collected, and a hemodialysis patient data set is obtained, wherein the hemodialysis patient data set includes features and corresponding feature values; The features are various attributes of the collected patients, and the feature values ​​are specific values ​​corresponding to the various attributes; S2: Preprocessing of hemodialysis patient dataset; The data preprocessing module preprocesses the data in the hemodialysis patient dataset by processing missing data and outliers, including the following sub-steps: S21: Handling missing data; Determine whether there is missing data in the hemodialysis patient data set, and if so, process the missing data; if not, proceed to step S22; S211: Determine whether there is missing data; Determine whether there is missing data in the hemodialysis patient dataset; Specific judgment methods include counting missing values ​​and visualizing data distribution; After the judgment is completed, the number of missing data is obtained through the sum function; If there is no missing data in the hemodialysis patient data set, proceed to step S22; if there is missing data, proceed to step S212; S212: Evaluate the impact of missing data; Assess the impact of missing data on the overall data distribution. Specific assessment methods include comparing statistical indicators, calculating correlations, and comparing data distributions. S213: Processing of missing data; Process the missing data according to the evaluation result of step S212. When the amount of missing data is small and the impact of the missing data on the overall data distribution is small, use the filling method or the deletion method. When the amount of missing data is large and the missing data has a great impact on the overall data distribution, deep learning methods and multiple interpolation methods are used to fill in the data; When there are missing data in the continuous data, interpolation is performed, and the interpolation method includes linear interpolation or polynomial interpolation; S22: Handling outliers; Determine whether there are abnormal values ​​in the hemodialysis patient data set. If there are abnormal values, process the abnormal values; if not, proceed to step S23; S221: Identify outliers; Identify whether there are outliers in the hemodialysis patient data set. Specific identification methods include box plot method and Z-score comparison detection method; S222: Processing abnormal values; Dealing with outliers by deleting or replacing them; Specifically, replacing outliers refers to replacing outliers with specific values ​​in the hemodialysis patient dataset, including medians and means; S3: Standardization of hemodialysis patient dataset; The data standardization module standardizes the data in the hemodialysis patient dataset; The diagnostic results of hemodialysis patients were standardized into mild malnutrition state, moderate malnutrition state, severe malnutrition state, and normal nutrition state; By calculating the difference value of each feature, the mean value and standard deviation of the difference value are obtained according to the calculated difference value of the feature, and the data is standardized by the Z-score standardization formula; S4: Feature selection and model building methods; The dimensionality reduction module combines data algorithms to identify features associated with malnutrition in the hemodialysis patient dataset; The following sub-steps are included: S41: Preliminary removal of features with small differences; The data in the hemodialysis patient data set are clustered and analyzed using a clustering algorithm to assign each sample to a corresponding cluster; the distribution of each feature in each cluster is calculated, and the distribution differences of each feature in different clusters are measured using variance; features with small differences are removed; S42: Data clustering calculation generation model; Generate a clustering model. The key indicators for evaluating the clustering model include silhouette coefficient, purity and accuracy of nutritional status within the cluster; S43: Randomly combine features for iterative clustering grouping; From the features initially screened in step S41, feature subsets of different combinations are randomly selected, and these subsets are used to re-cluster the data, and multiple iterations of cluster analysis are performed; S5: Model training and generation; Based on the feature selection and cluster analysis of S4, the malnutrition status is predicted through the final screened features and clustering model, including the following sub-steps: S51: prepare training data; The feature set is determined by the feature selection result of step S4, which is the training data; S52: Use clustering models to predict whether a patient is in a malnourished state; S53: Training using clustering model; S54: Model evaluation and optimization; S55: Use the trained clustering model to predict new data.

2. The method for predicting malnutrition in hemodialysis patients based on AI technology according to claim 1, characterized in that: In step S211, the method for determining whether there is missing data in the data set is as follows: Count missing values: Use the isnull or isna function to identify missing data in a dataset; Visualize data distribution: Draw charts to show the distribution of data in a dataset and identify missing data; After the judgment is completed, the missing data is summarized through the sum function to obtain the total number of missing data in the data set.

3. The method for predicting malnutrition in hemodialysis patients based on AI technology as claimed in claim 1, characterized in that: In step S212, statistical indicators are compared: according to the missing data detected in step S211, statistical indicators of the hemodialysis patient data set including the features corresponding to the missing data and statistical indicators of the hemodialysis patient data set without the features corresponding to the missing data are calculated, and the statistical indicators include mean, median, and standard deviation; the relative difference or percentage difference of the calculated statistical indicators is compared, and if the difference is small, it means that the missing data has a small impact on the overall data distribution; otherwise, it means that the missing data has a large impact on the overall data distribution; Calculate the correlation: Calculate the Pearson correlation coefficient between each feature in the hemodialysis patient data set through the corr function to obtain the correlation matrix, and extract the correlation coefficient between the feature corresponding to the missing data and other features in the correlation matrix. If the correlation coefficient is small, it means that the missing data has a small impact on the overall data distribution; otherwise, it means that the missing data has a large impact on the overall data distribution; Compare data distribution: Use a histogram to plot the data distribution of the hemodialysis patient data set containing the features corresponding to the missing data and the data distribution of the hemodialysis patient data set without the features corresponding to the missing data. Use a box plot to compare the data distribution in the above two cases. If the data distributions of the two are similar, the missing data has a small impact on the overall data distribution. Otherwise, the missing data has a large impact on the overall data distribution.

4. The method for predicting malnutrition in hemodialysis patients based on AI technology according to claim 1, characterized in that: In step S221, the box plot method: draw a box plot, wherein the box plot contains the median, upper quartile, lower quartile, upper limit, and lower limit of the data in the hemodialysis patient data set; If there is a data point in the box plot that is higher than the upper limit or lower than the lower limit, the data corresponding to the data point is an outlier, otherwise it is a non-outlier; Score comparison detection method: Calculate the mean and standard deviation of the data in the hemodialysis patient data set, and calculate the standard score Z-score of each data point based on the calculated mean, standard deviation and the data value in the hemodialysis patient data set through the Z-score standardization formula. Set a threshold in advance. If the standard score Z-score of the data point is greater than the threshold, it is an abnormal point, otherwise it is a non-abnormal point; the Z-score standardization formula is as follows: ； Among them, Z-score is the standard score, x is the data value in the hemodialysis patient data set, is the mean value and s is the standard deviation.

5. The method for predicting malnutrition in hemodialysis patients based on AI technology according to claim 1, characterized in that: In step S3, the data standardization module performs standardization processing on the data in the hemodialysis patient data set through data standardization; The hemodialysis patient dataset contains features of two types of tests: qualitative and quantitative; For the characteristics of qualitative detection, the detection result, i.e., the characteristic value, is represented by a numerical value, with a positive value of 1 and a negative value of 0; For the features of quantitative detection, the normal range of the detection results of this feature is determined in advance, the actual detection results are compared with the normal range, and the difference value of each feature is calculated; The normal range of the test result includes a single value and an interval; If the normal range of the test result is a single value, the difference value = |actual test result - normal range of the test result|; If the test result is an interval, calculate the center value of the interval, which is the average of the upper and lower limits of the interval, and the difference value = |actual test result-interval center value|; The mean and standard deviation of the difference values ​​are obtained based on the calculated difference values ​​of the features, and the data is standardized using the Z-score standardization formula; At this time, in the Z-score standardization formula, Z-score is the standard score, x is the difference value, is the mean of the difference values, and s is the standard deviation of the feature difference values.

6. The method for predicting malnutrition in hemodialysis patients based on AI technology according to claim 1, characterized in that: The specific contents of step S4 are as follows: S41: Preliminary removal of features with small differences; Cluster analysis is performed on the data in the hemodialysis patient dataset using a clustering algorithm; The clustering algorithms include k-means clustering, hierarchical clustering, and DBSCAN clustering; The data of each patient in the hemodialysis patient data set is recorded as a sample, and the above three clustering algorithms are run on each sample respectively. Each clustering algorithm assigns each sample to the corresponding cluster respectively; The method for setting the number of clusters for each clustering algorithm includes: the number of clusters for k-means clustering and hierarchical clustering uses the silhouette coefficient to determine the optimal number of clusters, and DBSCAN clustering does not require the number of clusters to be set in advance, but does require the eps and min_samples parameters to be set, which are determined by observing the data point distance graph; Calculate the distribution of each feature in each cluster, and measure the distribution difference of each feature in different clusters by variance; Calculate the variance of each feature contained in each cluster according to the variance formula, arrange the calculated variance of each feature in descending order, and draw a curve after the variance is sorted, in which the horizontal axis of the curve is the feature name and the vertical axis is the variance of the corresponding feature; The kneele algorithm is used to obtain the slope change of the curve, and the inflection point of the curve is detected according to the slope change. The features before the first inflection point of the detected curve show that there are large differences between different clusters, and different clusters are distinguished; The features that fail to show significant differences in different clusters in the data set, that is, the features after the first inflection point, are removed. The intersection of the features with significant differences obtained by the three clustering algorithms is the features after preliminary dimensionality reduction. The entire experimental process is iterative. The conditions for the end of the iteration are determined by the following three points: (1) The curve of feature variance ranking is stable: When the position of the first inflection point of the variance ranking curve calculated by the Kneedle algorithm no longer changes significantly in consecutive iterations, it indicates that the features that can distinguish different clusters have stabilized and the iteration is stopped; (2) Convergence of feature intersection of the three clustering algorithms: Compare the large difference feature sets obtained by K-means clustering, hierarchical clustering and DBSCAN. If the feature intersection gradually decreases after each iteration and the feature set of the final intersection no longer changes significantly, it means that a stable preliminary dimensionality reduction feature has been found and the iteration is stopped. (3) Clustering performance indicators meet expectations: The clustering effect after each iteration is evaluated using indicators such as silhouette coefficient, purity of nutrient status within the cluster, and accuracy; If these indicators do not improve significantly over multiple iterations, it means that the current feature set is effective enough and no further feature removal is needed; At least two of the above conditions are satisfied, indicating that the iterative process of step S41 has reached the state of optimal feature screening; S42: Data clustering calculation generation model; S41 initially removes features with small differences to obtain new features, and retains the new features for each cluster obtained by each clustering algorithm in S41. For each sample, a vote is conducted based on the clustering results of K-means, hierarchical clustering, and DBSCAN, and the most frequent cluster is selected as the final result; The cluster labels of the three algorithms are added to the dataset as new features to further improve the clustering effect; Labeling the nutritional status of each patient in each cluster; the nutritional status includes normal nutrition and malnutrition; The malnutrition status labels are all set to 1, and the malnutrition status includes mild, moderate, and severe malnutrition; The label for normal nutritional status is set to 0; For patients in each cluster, statistics were taken based on their nutritional status, and the proportion of people in each nutritional status in the cluster was calculated; Label each cluster with a nutritional status label, which is the nutritional status with the highest proportion in each cluster; Finally, a clustering model was generated. The key indicators for evaluating the clustering model included the silhouette coefficient, the purity and accuracy of the nutritional status within the cluster. If the silhouette coefficient was above 0.5, the purity of the nutritional status within the cluster was above 80%, and the accuracy was above 85%, it indicated that the clustering model could be used to identify the nutritional status of new data. S43: Randomly combine features for iterative clustering grouping; From the features initially screened in step S41, feature subsets of different combinations are randomly selected and used to re-cluster the data; in each clustering, the k-means clustering, hierarchical clustering and DBSCAN clustering methods are re-applied, the previous cluster number determination and clustering steps are repeated, and the results are fused; In each iteration, the same process as S42 was maintained, that is, a nutritional status label was assigned to each cluster and the nutritional status ratio in each cluster was calculated; After each clustering, the effect of the clustering model was evaluated to ensure that the model could effectively distinguish different nutritional states. The silhouette coefficient of each cluster, the purity and accuracy of the nutritional state within the cluster were calculated to ensure that the clustering results could effectively distinguish the nutritional states. When similar feature combinations in multiple iterations repeatedly obtain similar clustering results, and the indicators such as the purity of the nutritional status within the cluster, the silhouette coefficient and the accuracy rate reach the set threshold, that is, when the silhouette coefficient remains stable above 0.5, the purity value remains above 80%, and the accuracy rate remains above 85%, and there is no significant improvement in further iterations, the iteration is stopped; Characteristics found to be highly associated with malnutrition in prior studies remained constant throughout the analysis.

7. The method for predicting malnutrition in hemodialysis patients based on AI technology according to claim 1, characterized in that: The specific contents of step S5 are as follows: The nutritional status prediction module is based on the feature selection and cluster analysis of S4. It predicts the malnutrition status through the final screened features and clustering model, including the following sub-steps: S51: prepare training data; The feature selection results of step S4 are used to determine the feature set. These features are the most effective features for distinguishing malnutrition status after multiple clustering iterations. Each sample has been labeled with a nutritional status label, with mild, moderate, and severe malnutrition as 1 and normal nutrition as 0. S52: Use clustering models to predict whether a patient is in a malnourished state; Patient samples have been assigned to different clusters through the clustering model, and a nutritional status label has been assigned to each cluster. This clustering structure is used to build a prediction model; S53: Training using clustering model; Use the filtered feature set as input data; According to the characteristics of each new sample, the new sample is assigned to a cluster using the trained clustering model, and the The calculated nutritional status ratio is used to assign predicted nutritional status labels to new samples; Specifically, new samples will be assigned to a cluster, the nutritional status ratio in the cluster determines the label of the sample, and the prediction accuracy is evaluated using the known nutritional status labels. The performance of the model is evaluated by accuracy, precision, recall, F1 score and other criteria; S54: Model evaluation and optimization; Evaluation indicators include accuracy, precision, recall, and F1 score; if the initial training results are not ideal, adjust the parameters of the clustering algorithm or recombine features during feature selection; S55: Use the trained clustering model to predict new data; For each new patient, after inputting its characteristics, the clustering model assigns the patient to a cluster and predicts whether the patient is in a malnourished state based on the nutritional status label of the cluster.

8. A system for predicting malnutrition in hemodialysis patients based on AI technology, used to implement the method for predicting malnutrition in hemodialysis patients based on AI technology according to any one of claims 1 to 7, characterized in that: It includes data collection module, data preprocessing module, data standardization module, dimension reduction module and nutritional status prediction module; The data collection module is used to collect data related to hemodialysis patients; The data preprocessing module is used to preprocess the data in the hemodialysis patient data set; The data standardization module is used to perform standardization processing on the data in the hemodialysis patient data set; The dimension reduction module is used to perform cluster analysis on the data in the hemodialysis patient data set, remove features with little correlation with malnutrition, and perform iterative clustering and grouping to obtain features that are highly correlated with malnutrition; The nutritional status prediction module is used to establish a prediction model and predict whether the patient is in a malnutrition state.

Images