Intelligent health monitoring system for patient rehabilitation nursing

Abstract

The invention discloses an intelligent health monitoring system for patient rehabilitation nursing, which belongs to the technical field of medical data processing and comprises a medical data acquisition module, a medical data preprocessing module, an intelligent health monitoring model building module, an intelligent health monitoring model parameter optimizing module and a patient rehabilitation nursing module. The method specifically comprises the following steps: constructing a candidate data set based on the selection number of each type of data labels, introducing a neighborhood label coupling degree and a neighborhood data deviation index to design a noise judgment coefficient, identifying non-noise data and constructing a training subset, introducing a relative error and an adjustment factor to update a data weight, and carrying out weighted combination on weak classifiers; according to the method, chaotic mapping and a sine function are combined, random disturbance terms obeying normal distribution are added, initialization of individual positions is completed, the individual positions are updated based on judgment values and global optimal positions, the health condition of a patient is accurately judged, and therefore the rehabilitation nursing quality and effect of the patient are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical data processing, and specifically refers to a health intelligent monitoring system for patient rehabilitation care. Background Art

[0002] The health intelligent monitoring system uses artificial intelligence technology to collect and analyze patients' medical data in real time, monitor and evaluate patients' health status, support personalized rehabilitation care plans, and improve the efficiency of rehabilitation care. However, in the existing health intelligent monitoring system, there is an imbalance in medical data of different categories of patients, and the accuracy limitation of measuring devices will generate measurement errors, resulting in noise in the collected patients' medical data, leading to the problem that the health conditions of patients cannot be accurately judged; there are large individual differences among different patients in the existing health intelligent monitoring system, and the medical data of patients is complex and uncertain, resulting in the problem that the important medical characteristics of different patients cannot be accurately captured, and misjudgment of the health conditions of patients occurs. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a health intelligent monitoring system for patient rehabilitation care. Aiming at the problem that in the existing health intelligent monitoring system, there is an imbalance in medical data of different categories of patients, and the accuracy limitation of measuring devices will generate measurement errors, resulting in noise in the collected patients' medical data, leading to the problem that the health conditions of patients cannot be accurately judged, this solution constructs a candidate data set based on the selection quantity of each type of data label, and can comprehensively learn the medical data characteristics in various health states; introduces the neighborhood label coupling degree and the neighborhood data deviation index to design a noise determination coefficient, identifies non-noise data and constructs a training subset, and more accurately learns the true characteristics of patients' medical data; trains weak classifiers and calculates weights, introduces relative error and adjustment factors to update data weights, and performs weighted combination on the weak classifiers to obtain a health intelligent monitoring model, accurately monitors the health status of patients, and provides a more reliable basis for the adjustment of rehabilitation care plans; aiming at the problem that in the existing health intelligent monitoring system, there are large individual differences among different patients, and the medical data of patients is complex and uncertain, resulting in the problem that the important medical characteristics of different patients cannot be accurately captured, and misjudgment of the health conditions of patients occurs, this solution combines chaotic mapping and sine function, and then adds a random perturbation term subject to normal distribution to complete the initialization of individual positions, and better deals with the random fluctuations and uncertainties in patients' medical data; updates individual positions based on judgment values and the global optimal position, determines optimal parameters, and customizes model parameters suitable for each patient's own situation to accurately capture the important medical characteristics of patients, so as to more accurately judge the health conditions of patients, thereby improving the quality and effect of patient rehabilitation care.

[0004] A health intelligent monitoring system for patient rehabilitation care provided by the present invention includes a medical data collection module, a medical data preprocessing module, a health intelligent monitoring model construction module, a health intelligent monitoring model parameter optimization module, and a patient rehabilitation care module;

[0005] The medical data collection module collects historical patient medical data;

[0006] The medical data preprocessing module performs data cleaning, data conversion, data normalization, and dataset construction processing on the collected patient medical data;

[0007] The health intelligent monitoring model construction module constructs a candidate dataset based on the selected number of each type of data label, introduces the neighborhood label coupling degree and the neighborhood data deviation index to design a noise determination coefficient, identifies non-noise data and constructs a training subset, trains weak classifiers and calculates weights, introduces the relative error and a regulation factor to update the data weights, and performs weighted combination on the weak classifiers to obtain a health intelligent monitoring model;

[0008] The health intelligent monitoring model parameter optimization module combines the chaotic mapping and the sine function, adds a random perturbation term that follows a normal distribution, completes the initialization of the individual position, updates the individual position based on the judgment value and the global optimal position, and determines the optimal parameters;

[0009] The patient rehabilitation care module collects real-time patient medical data, and based on the health status output by the health intelligent monitoring model, medical staff adjusts the patient's rehabilitation care plan.

[0010] Further, the medical data collection module collects historical patient medical data; the historical patient medical data includes vital sign data, activity data, life data, pain level data, and health status; the health status is used as a data label.

[0011] Further, the medical data preprocessing module performs data cleaning, data conversion, data normalization, and dataset construction processing on the collected patient medical data; the data cleaning includes handling missing values, outliers, and duplicate values; the data conversion is to convert the data into a vector form; the data normalization is to unify the data range based on the maximum-minimum normalization method; the dataset construction is to construct a training dataset and a test dataset based on the patient medical data after data cleaning, data conversion, and data normalization.

[0012] Further, the health intelligent monitoring model construction module is provided with an initial data weight unit, a candidate dataset design unit, a training subset design unit, a weak classifier training unit, a weak classifier weight calculation unit, a data weight update unit, and a weighted combination unit, and specifically includes the following contents:

[0013] An initial data weight unit that initializes the weight value of each patient's medical data in the training dataset to ; where F is the number of patient medical data in the training dataset;

[0014] A designed candidate dataset unit that pre-sets the maximum number of training times to T. The candidate dataset for the first training consists of n 1 patient medical data randomly selected from the training dataset. For the second to the T-th training, the construction of the candidate dataset includes the following:

[0015] Calculate the selection quantity of each class of data label, and the formula used is as follows:

[0016] ;

[0017] In the formula, is the selection quantity of the i-th class of data label in the t-th training, a i is the number of patient medical data belonging to the i-th class of data label in the training dataset, is the number of patient medical data that were misclassified in the weak classifier of the (t - 1)-th training and belong to the i-th class of data label. t is the training times index, i is the data label class index, n 2 is the maximum number of patient medical data in the candidate dataset, δ is the selection factor, is rounding down;

[0018] Construct the candidate dataset, and the formula used is as follows:

[0019] ;

[0020] In the formula, h t and h t-1 are the candidate datasets for the t-th and (t - 1)-th trainings respectively, is the union, l i is the i-th class of data label, L is the set of data label classes, A i is the set constructed from the patient medical data belonging to the i-th class of data label in the training dataset, is the set constructed from the patient medical data that were misclassified in the weak classifier of the (t - 1)-th training and belong to the i-th class of data label, is randomly selected from the difference set of A i and patient medical data; is to perform i operations on each class of data label l in the class set L, and then take the union of the operation results corresponding to each class of data label l i ;​

[0021] Design a training subset unit, including the following:

[0022] Design the neighborhood label coupling degree, calculate the k nearest neighbor data of each patient's medical data in the candidate dataset, and construct the first k-nearest neighbor set , and then calculate the neighborhood label coupling degree of each patient's medical data. The formula used is as follows:

[0023] ;

[0024] In the formula, x j and x v are the medical data of the j-th and v-th patients respectively, , and are the first k-nearest neighbor set, neighborhood label coupling degree, and data label of x t in h j respectively, is the data label of x v , j and v are patient medical data indices, is the coupling factor; is the indicator function, which is 1 if and 0 otherwise;

[0025] Design the neighborhood data deviation index, calculate the k nearest neighbor data of each patient's medical data in each category in the candidate dataset, and construct the second k-nearest neighbor set , and then calculate the neighborhood data deviation index of each patient's medical data. The formula used is as follows:

[0026] ;

[0027] In the formula, is the neighborhood data deviation index of x t in h j on l i , exp(•) is the exponential function, is the second k-nearest neighbor set of x t in h j on l i , , is the Euclidean distance between x j and x f , and are the average Euclidean distances of x j and x b in the second k-nearest neighbor set on l i , x b and x fThey are the medical data of the b-th and f-th patients respectively, where b and f are the indexes of the medical data of patients;

[0028] Design the noise determination coefficient, and the formula used is as follows:

[0029] ;

[0030] In the formula, is the noise determination coefficient of x in h t where x j ; is the data label of x j ; is the neighborhood data deviation index of x in h t where x j on , and ρ is the control factor; is to find the maximum neighborhood data deviation index of x in h among all the data labels l i belonging to t where x j on l i ; ;

[0031] Construct a training subset, and preset a noise threshold . If , then x in h t is noise data; otherwise, x in h j is non-noise data; construct the training subset t for the t-th training based on the non-noise data in h j ; t ;

[0032] Train the weak classifier unit, use the weights of the current patient's medical data, and train the training subset based on the Naive Bayes classifier to obtain the weak classifier H t for the t-th training;

[0033] Calculate the weak classifier weight unit, and the formula used is as follows:

[0034]

[0035] ; In the formula, α t is the weight of H t , is the error rate of H t ;

[0036] Update the data weight unit, introduce the relative error and the adjustment factor to update the weights of the patient's medical data, and the formula used is as follows:

[0037] ;

[0038] In the formula, and are the weights of x at the (t + 1)-th and t-th trainings respectively, x j is the medical data of the g-th patient, and g is the index of the patient's medical data. g is the weight of x at the t-th training, g ; is the classification label of H t for x j ; is the relative error of H t on x j ; and γ are the update factor and the adjustment factor respectively. μ t is the average value of all t of H on the training subset, and σ t is the standard deviation of all t of H on the training subset;

[0039] Weighted combination unit. If the maximum number of trainings is reached, the T weak classifiers obtained from T trainings are weighted and combined to obtain a health intelligent monitoring model; otherwise, the number of trainings is incremented by 1 and the design candidate dataset unit is returned to continue training.

[0040] Furthermore, the health intelligent monitoring model parameter optimization module is provided with an initial individual position unit, an updated individual position unit, and an optimal parameter determination unit, and specifically includes the following:

[0041] Initial individual position unit. A parameter search space is established for the selection factor δ, coupling factor , control factor ρ, update factor , and adjustment factor γ in the health intelligent monitoring model. R initial individual positions are generated within the parameter search space. The individual positions are used as representatives of the parameters of the health intelligent monitoring model, and the cross-entropy loss of the health intelligent monitoring model trained based on the parameters on the test dataset is used as the fitness value of the individual positions. By combining the chaotic mapping and the sine function and adding a random perturbation term that follows a normal distribution, the formula for the initial individual position is as follows:

[0042] ;

[0043] In the formula, and are the (r + 1)-th and r-th initial individual positions respectively, r is the individual position index, sin(•) is the sine function, z is the chaotic coefficient, is a random disturbance term that follows a normal distribution with a mean of 0 and a variance of σ 2 ; is the disturbance coefficient;

[0044] Update the individual position unit. Generate a judgment value randomly for each individual position, and update the individual position based on the judgment value and the global optimal position. The formula used is as follows:

[0045] ;

[0046] In the formula, and are the r-th individual positions at the (p + 1)-th and p-th searches respectively. p is the search number index, and v is the individual position index. and are the global optimal positions of the r-th and v-th individual positions since the p-th search respectively. is the global optimal position among the R individual positions at the p-th search. is the judgment value of the r-th individual position at the p-th search, and η, λ, and are three random numbers that do not interfere with each other;

[0047] Determine the optimal parameter unit. Preset the fitness threshold and the maximum number of searches. Update the fitness value of the individual position. When the fitness value of an individual position X is lower than the fitness threshold, the model parameters represented by the individual position X are the optimal parameters, and a health intelligent monitoring model is constructed based on the optimal parameters; otherwise, if the maximum number of searches is reached, return to the initial individual position unit; otherwise, increment the search number by 1 and return to the updated individual position unit to continue the search.

[0048] Furthermore, the patient rehabilitation care module collects real-time patient medical data, which includes vital sign data, activity data, life data, and pain level data. After preprocessing the real-time patient medical data, it is input into the health intelligent monitoring model constructed based on the optimal parameters for classification to obtain the patient's health status, and medical staff adjust the rehabilitation care plan according to the patient's health status.

[0049] The beneficial effects achieved by the present invention using the above solution are as follows:

[0050] (1) Aiming at the problems in the existing health intelligent monitoring system that there is an imbalance in the medical data of different types of patients, and the accuracy limitation of the measurement device will generate measurement errors, resulting in noise in the collected patient medical data, which makes it impossible to accurately judge the health status of patients. This solution constructs a candidate data set based on the selection quantity of each type of data label, which can comprehensively learn the characteristics of medical data in various health states. The balanced data helps to accurately judge the health status of patients in different situations. By introducing the neighborhood label coupling degree and the neighborhood data deviation index to design the noise determination coefficient, non-noise data is identified and a training subset is constructed, so as to more accurately learn the true characteristics of patient medical data, thereby improving the accuracy of judging the health status of patients. Weak classifiers are trained and weights are calculated. By introducing the relative error and the adjustment factor to update the data weights and performing weighted combination on the weak classifiers, a health intelligent monitoring model is obtained, which is more adaptable to the dynamic changes of patient medical data and can accurately monitor the health status of patients, providing a more reliable basis for the adjustment of the rehabilitation nursing plan.

[0051] (2) Aiming at the problems in the existing health intelligent monitoring system that there are large individual differences among different patients, and the patient medical data has complexity and uncertainty, resulting in the inability to accurately capture the important medical characteristics of different patients and misjudging the health status of patients. This solution uses the individual position as the parameter representative of the health intelligent monitoring model, combines the chaotic mapping and the sine function, and then adds a random perturbation term that follows the normal distribution to complete the initialization of the individual position, expanding the coverage range in the initial stage and better coping with the random fluctuations and uncertainties in patient medical data. Based on the judgment value and the global optimal position, the individual position is updated to determine the optimal parameters, and model parameters suitable for each patient's own situation are customized, improving the adaptability to different patients, accurately capturing the important medical characteristics of patients, and thus more precisely judging the health status of patients, thereby improving the quality and effect of patient rehabilitation nursing. Brief Description of the Drawings

[0052] Figure 1 It is a schematic diagram of a health intelligent monitoring system for patient rehabilitation nursing provided by the present invention;

[0053] Figure 2 It is a schematic diagram of the module for constructing a health intelligent monitoring model;

[0054] Figure 3 It is a schematic diagram of the module for optimizing the parameters of the health intelligent monitoring model.

[0055] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. Detailed Embodiments

[0056] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0057] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0058] Embodiment 1. Refer to Figure 1 , a health intelligent monitoring system for patient rehabilitation care provided by the present invention includes a medical data acquisition module, a medical data preprocessing module, a health intelligent monitoring model construction module, a health intelligent monitoring model parameter optimization module, and a patient rehabilitation care module;

[0059] The medical data acquisition module collects historical patient medical data and sends the data to the medical data preprocessing module;

[0060] The medical data preprocessing module receives the data sent by the medical data acquisition module, performs data cleaning, data conversion, data normalization, and dataset construction processing on the collected patient medical data, and sends the data to the health intelligent monitoring model construction module;

[0061] The health intelligent monitoring model construction module receives the data sent by the medical data preprocessing module, constructs a candidate dataset based on the selection quantity of each type of data label, introduces the neighborhood label coupling degree and the neighborhood data deviation index to design a noise determination coefficient, identifies non-noise data and constructs a training subset, trains weak classifiers and calculates weights, introduces the relative error and adjustment factor to update the data weights, performs weighted combination on the weak classifiers to obtain a health intelligent monitoring model, and sends the data to the health intelligent monitoring model parameter optimization module;

[0062] The health intelligent monitoring model parameter optimization module receives the data sent by the health intelligent monitoring model construction module, combines the chaotic mapping and the sine function, then adds a random perturbation term obeying the normal distribution to complete the initialization of the individual position, updates the individual position based on the judgment value and the global optimal position, determines the optimal parameters, and sends the data to the patient rehabilitation care module;

[0063] The patient rehabilitation nursing module receives the data sent by the health intelligent monitoring model parameter optimization module, collects real-time patient medical data, and based on the health status output by the health intelligent monitoring model, medical staff adjust the patient's rehabilitation nursing plan.

[0064] Example 2, refer to Figure 1 , based on the above example, in the medical data collection module, the historical patient medical data includes vital sign data, activity data, living data, pain level data, and health status; the vital sign data includes heart rate, blood oxygen saturation, body temperature, respiratory rate, blood pressure, and blood glucose; the activity data includes the number of steps, activity intensity, and body position changes; the living data includes diet, work and rest time, and sleep quality; the health status includes normal status, mild abnormal status, moderate abnormal status, severe abnormal status, and critical status, and the health status is used as the data label.

[0065] Example 3, refer to Figure 1 , based on the above example, in the medical data preprocessing module, data cleaning includes handling missing values, outliers, and duplicate values; data conversion is to convert the data into vector form; data normalization is to unify the data range based on the maximum-minimum normalization method; constructing a data set is to construct a training data set and a test data set based on the patient medical data after data cleaning, data conversion, and data normalization.

[0066] Example 4, refer to Figure 1 and Figure 2 , based on the above example, in the module for constructing a health intelligent monitoring model, there are an initial data weight unit, a designed candidate data set unit, a designed training subset unit, a trained weak classifier unit, a calculated weak classifier weight unit, an updated data weight unit, and a weighted combination unit, specifically including the following:

[0067] Initial data weight unit, set the initial value of the weight of each patient medical data in the training data set to ; where F is the number of patient medical data in the training data set;

[0068] Designed candidate data set unit, dynamically adjust the composition of the candidate data set according to the progress of model training, so that the model can better learn the characteristics of various patient medical data, especially pay attention to the patient medical data that was misclassified before, and improve the accuracy of the model. Specifically: preset the maximum number of training times as T, the candidate data set for the first training consists of n 1 patient medical data randomly selected from the training data set. For the second to the Tth training, the construction of the candidate data set includes the following:

[0069] Calculate the selection quantity of each type of data label, and the formula used is as follows:

[0070] ;

[0071] wherein, is the number of selections of the data label of the i-th class at the t-th training, and a i is the number of medical data of patients belonging to the data label of the i-th class in the training dataset, is the number of medical data of patients that are misclassified and belong to the data label of the i-th class in the weak classifier at the (t - 1)-th training, t is the training times index, i is the data label class index, and n 2 is the maximum number of medical data of patients in the candidate dataset, and δ is a selection factor within the range of (0, 1), is the floor function;

[0072] Construct a candidate dataset, and the formula used is as follows:

[0073] ;

[0074] wherein, h t and h t-1 are the candidate datasets at the t-th and (t - 1)-th trainings respectively, is the union, l i is the data label of the i-th class, L is the set of data label classes, and A i is the set constructed from the medical data of patients belonging to the data label of the i-th class in the training dataset, is the set constructed from the medical data of patients that are misclassified and belong to the data label of the i-th class in the weak classifier at the (t - 1)-th training, is randomly selected from the difference set of A i and medical data of patients; is to perform i for each data label l in the class set L, and then take the union of the operation results corresponding to each data label l i ;

[0075] medical data of patients;

[0076] Design a training subset unit to extract the non-noise data in the candidate dataset as the training subset, effectively removing the interference of noise data on model training and improving the stability and accuracy of the model, including the following:

[0077] Design the neighborhood label coupling degree, calculate the k nearest neighbor data of each medical data of patients in the candidate dataset, and construct the first k-nearest neighbor set , and then calculate the neighborhood label coupling degree of each medical data of patients, and the formula used is as follows:​

[0078] ;

[0079] wherein, x j and x v are the medical data of the j-th and v-th patients respectively, , and are the first k-nearest neighbor set, neighborhood label coupling degree, and data label of x t in h j respectively, is the data label of x v , j and v are patient medical data indices, is a coupling factor in the range of (0, 1); is an indicator function, which is 1 if and 0 otherwise;

[0080] Design a neighborhood data deviation index, calculate the k nearest neighbor data of each patient's medical data in each category in the candidate dataset, and construct a second k-nearest neighbor set , and then calculate the neighborhood data deviation index of each patient's medical data. The formula used is as follows:

[0081] ;

[0082] wherein, is the neighborhood data deviation index of x t in h j on l i , exp(•) is an exponential function, is the second k-nearest neighbor set of x t in h j on l i , , is the Euclidean distance between x j and x f , and are the average Euclidean distances of x j and x b in the second k-nearest neighbor set on l i , x b and x f are the medical data of the b-th and f-th patients respectively, and b and f are patient medical data indices;

[0083] Design a noise determination coefficient. The formula used is as follows:

[0084] ;

[0085] wherein, is h t in x j the noise determination coefficient, is x j the data label, is h t in x j at the neighborhood data deviation index on, ρ is a control factor in the range of (0, 1); is in the ones belonging to all data labels l i find h t in x j in l i the maximum neighborhood data deviation index on ;

[0086] Construct a training subset, and preset a noise threshold If then h t the x in j is noise data; otherwise, h t the x in j is non-noise data; Based on the non-noise data in h t construct the training subset for the t-th training ;

[0087] Train the weak classifier unit, using the weights of the current patient's medical data, and train the training subset based on the Naive Bayes classifier to obtain the weak classifier H t for the t-th training;

[0088] Calculate the weak classifier weight unit, and the formula used is as follows:

[0089] ;

[0090] In the formula, α t is the weight of H t is the error rate of H t , D is the number of patient medical data in the training subset for the t-th training, ω j is the weight of x j is the classification label of H t for x j ; is the indicator function, if then it is 1, otherwise it is 0;

[0091] ​​​Update the data weight unit, introduce relative error and adjustment factor to update the weight of patient medical data, the formula used is as follows:

[0092] ;

[0093] In the formula, and are x at the t+1th and tth training times respectively. j The weight of x g is the medical data of the gth patient, g is the index of the patient's medical data, is the tth training time x g The weight of Yes H t x j The classification label of Yes H t In x j The relative error on and γ are update factors and adjustment factors in the range of (0, 1), μ t Yes H t All on the training subset The average value, σ t Yes H t All on the training subset The standard deviation of

[0094] The weighted combination unit, if the maximum number of training times is reached, performs weighted combination on the T weak classifiers obtained from T training times to obtain a health intelligent monitoring model; otherwise, the number of training times is increased by 1 and the unit returns to the design candidate data set to continue training; the formula used is as follows:

[0095] ;

[0096] In the formula, It is a health intelligent monitoring model.

[0097] By performing the above operations, aiming at the problems in the existing health intelligent monitoring system that there are imbalances in medical data of different categories of patients, and the accuracy limitations of measurement devices will generate measurement errors, resulting in noisy patient medical data collected, which makes it impossible to accurately judge the health status of patients. This solution constructs a candidate data set based on the number of selected data labels for each category, and can comprehensively learn the characteristics of medical data in various health states. The balanced data helps to accurately judge the health status of patients in different situations; introduces the neighborhood label coupling degree and neighborhood data deviation index to design a noise determination coefficient, identifies non-noise data and constructs a training subset, and more accurately learns the true characteristics of patient medical data, thereby improving the accuracy of judging the health status of patients; trains weak classifiers and calculates weights, introduces relative error and adjustment factors to update data weights, and performs weighted combination on weak classifiers to obtain a health intelligent monitoring model, which is more adaptable to the dynamic changes of patient medical data, accurately monitors the health status of patients, and provides a more reliable basis for the adjustment of rehabilitation nursing plans.

[0098] Example Five. Refer to Figure 1 and Figure 3 , this example is based on the above example. In the health intelligent monitoring model parameter optimization module, there are an initial individual position unit, an updated individual position unit, and a unit for determining optimal parameters, which specifically include the following:

[0099] The initial individual position unit establishes a parameter search space for the selection factor δ, coupling factor , control factor ρ, update factor , and adjustment factor γ in the health intelligent monitoring model. Initial R individual positions within the parameter search space, and use the individual positions as the parameter representatives of the health intelligent monitoring model. Take the cross-entropy loss of the health intelligent monitoring model trained based on the parameters for the test data set as the fitness value of the individual positions; combine the chaotic mapping and the sine function, and then add a random perturbation term that follows a normal distribution. The formula for the initial individual position is as follows:

[0100] ;

[0101] In the formula, and are the initial (r + 1)-th and r-th individual positions respectively, r is the individual position index, sin(•) is the sine function, z is the chaotic coefficient within the range of [0, 4], is a random perturbation term that follows a normal distribution with a mean of 0 and a variance of σ 2 , is the perturbation coefficient within the range of [0, 0.1];

[0102] Update the individual position unit. Generate a judgment value within the range of (0, 1) randomly for each individual position, and update the individual position based on the judgment value and the global optimal position, making the search process more flexible. It can be reasonably adjusted according to the relationship between the individual position and the global optimal position at different search stages, improving the search efficiency and avoiding premature convergence to the local optimal solution in the search process, so as to be more likely to find a better combination of model parameters. The formula used is as follows:

[0103] ;

[0104] In the formula, and are the r-th individual positions at the (p + 1)-th and p-th searches respectively. p is the search number index, and v is the individual position index. and are the global optimal positions of the r-th and v-th individual positions since the p-th search respectively. is the global optimal position among the R individual positions at the p-th search. The global optimal position is the individual position with the smallest fitness value. is the judgment value of the r-th individual position at the p-th search. η, λ, and are three non-interfering random numbers within the range of (0, 1).

[0105] Determine the optimal parameter unit. Preset the fitness threshold and the maximum number of searches, and update the fitness value of the individual position. When the fitness value of an individual position X is lower than the fitness threshold, the model parameters represented by the individual position X are the optimal parameters, and a health intelligent monitoring model is constructed based on the optimal parameters; otherwise, if the maximum number of searches is reached, return to the initial individual position unit; otherwise, increment the search number by 1 and return to the updated individual position unit to continue the search.

[0106] By performing the above operations, aiming at the problem in the existing health intelligent monitoring system that there are large individual differences among different patients, and the medical data of patients is complex and uncertain, resulting in the inability to accurately capture the important medical characteristics of different patients and misjudging the health status of patients. In this solution, the individual position is used as the parameter representative of the health intelligent monitoring model. The chaos mapping and the sine function are combined, and then a random perturbation term subject to a normal distribution is added to complete the initialization of the individual position, expanding the coverage range in the initial stage and better coping with the random fluctuations and uncertainties in the patient's medical data; update the individual position based on the judgment value and the global optimal position, determine the optimal parameters, customize the model parameters suitable for each patient's own situation, improve the adaptability to different patients, accurately capture the important medical characteristics of patients, and thus more accurately judge the health status of patients, thereby improving the quality and effect of patient rehabilitation care.

[0107] Example Six, refer toFigure 1 , this embodiment is based on the above-mentioned embodiment. In the patient rehabilitation care module, real-time patient medical data is collected. The real-time patient medical data includes vital sign data, activity data, living data, and pain level data. After preprocessing the real-time patient medical data, it is input into a health intelligent monitoring model constructed based on optimal parameters for classification to obtain the patient's health status, and medical staff adjust the rehabilitation care plan according to the patient's health status.

[0108] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0109] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0110] The above describes the present invention and its implementation manners. Such a description is not restrictive, and what is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and design similar structural manners and embodiments to this technical solution without creative efforts without departing from the purpose of the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A health intelligent monitoring system for patient rehabilitation care, characterized in that: It includes medical data collection module, medical data preprocessing module, health intelligent monitoring model building module, health intelligent monitoring model parameter optimization module and patient rehabilitation nursing module; The medical data collection module is used to collect historical patient medical data; The medical data preprocessing module performs data cleaning, data conversion, data normalization and data set construction on the collected patient medical data; The module for constructing a health intelligent monitoring model constructs a candidate data set based on the number of selected data labels of each type, introduces the neighborhood label coupling degree and the neighborhood data deviation index to design the noise determination coefficient, identifies non-noise data and constructs a training subset, trains weak classifiers and calculates weights, introduces relative errors and adjustment factors to update data weights, and performs weighted combination on weak classifiers to obtain a health intelligent monitoring model; The parameter optimization module of the health intelligent monitoring model combines the chaotic map and the sine function, and then adds a random disturbance term that obeys the normal distribution to complete the initialization of the individual position, and updates the individual position based on the judgment value and the global optimal position to determine the optimal parameters; The patient rehabilitation nursing module collects real-time patient medical data, and based on the health status output by the health intelligent monitoring model, medical staff adjust the patient's rehabilitation nursing plan; The module for constructing a health intelligent monitoring model is provided with a data weight updating unit, which introduces relative errors and adjustment factors to update the weights of the patient's medical data. The formula used is as follows: ； In the formula, and are x at the t+1th and tth training times respectively. j The weight of is the tth training time x g The weight of x j and x g are the medical data of the jth and gth patients respectively, j and g are the indexes of the patient medical data, t is the index of the number of training times, is the training subset for the t-th training, Yes H t The error rate, H t is the weak classifier trained for the tth time, Yes H t x j The classification label of is x j The data labels, Yes H t In x j The relative error on and γ are the update factor and adjustment factor respectively, μ t Yes H t All on the training subset The average value, σ t Yes H t All on the training subset The standard deviation of .

2. The intelligent health monitoring system for patient rehabilitation care according to claim 1, characterized in that: The module for constructing a health intelligent monitoring model includes an initial data weight unit, a candidate data set design unit, a training subset design unit, a weak classifier training unit, a weak classifier weight calculation unit, a data weight update unit, and a weighted combination unit, which specifically include the following contents: The initial data weight unit sets the initial weight value of each patient's medical data in the training data set to ; Where F is the number of patient medical data in the training dataset; Design the candidate data set unit, pre-set the maximum number of training times to T, the candidate data set for the first training is composed of n1 patient medical data randomly selected from the training data set, and for the second to Tth training, the construction of the candidate data set includes the following: Calculate the number of choices for each data label using the following formula: ； In the formula, is the number of selected data labels of the i-th category during the t-th training, a i is the number of patient medical data belonging to the i-th category data label in the training dataset, is the number of patient medical data that are misclassified in the weak classifier trained for the t-1th time and belong to the i-th category data label, i is the category index of the data label, n2 is the maximum number of patient medical data in the candidate data set, δ is the selection factor, is rounded down; The candidate data set is constructed using the following formula: ； In the formula, h t and h t-1 are the candidate data sets for the t-th and t-1-th training respectively, is a union, l i is the i-th data label, L is the category set of data labels, A i is a set constructed from the patient medical data belonging to the i-th category data label in the training data set. is a set constructed from the patient medical data that is misclassified in the weak classifier trained for the t-1th time and belongs to the i-th category data label, It is from A i and Randomly select from the difference Individual patient medical data; is the label l of each data type in the category set L i implement After the operation, each type of data label l i The corresponding operation results are taken as the union; Design training subset units; Train the weak classifier unit, using the weights of the current patient's medical data, based on the naive Bayes classifier on the training subset Train and get the weak classifier H of the tth training t ; Calculate the weak classifier weight unit, the formula used is as follows: ； In the formula, α t Yes H t The weight of Yes H t The error rate; Update data weight unit; The weighted combination unit, if the maximum number of training times is reached, performs weighted combination on the T weak classifiers obtained from T training times to obtain a health intelligent monitoring model; otherwise, the number of training times is increased by 1 and returns to the design candidate data set unit to continue training.

3. The intelligent health monitoring system for patient rehabilitation care according to claim 2 is characterized in that: The design training subset unit specifically includes the following contents: Design the neighborhood label coupling degree, calculate the k nearest neighbor data of each patient's medical data in the candidate data set, and build the first k nearest neighbor set , and then calculate the neighborhood label coupling degree of each patient's medical data. The formula used is as follows: ； In the formula, x j and x v are the medical data of the jth and vth patients respectively, , and They are h t Medium j The first k nearest neighbor set, neighborhood label coupling degree and data label, is x v The data label of j and v is the patient medical data index. is the coupling factor; is an indicator function, if If yes, it is 1, otherwise it is 0; Design the neighborhood data deviation index, calculate the k nearest neighbor data of each patient's medical data in each category in the candidate data set, and build the second k nearest neighbor set , and then calculate the neighborhood data deviation index of each patient's medical data. The formula used is as follows: ； In the formula, Yes t Medium j In l i The neighborhood data on deviates exponentially, exp(•) is an exponential function, Yes t Medium j In l i The second k nearest neighbor set on , , is x j and x f The Euclidean distance between and They are x j and x b In l i The average Euclidean distance in the second k nearest neighbor set, x b and x f are the bth and fth patient medical data respectively, b and f are the patient medical data indexes; Design the noise determination coefficient, the formula used is as follows: ； In the formula, Yes t Medium j The noise determination coefficient is is x j The data labels, Yes t Medium j exist The neighborhood data on deviates from the exponent, and ρ is the control factor; is in All data labels l i In the t Medium j In l i The largest neighborhood data deviation index ; Construct a training subset and pre-set the noise threshold ,like , then h t x in j is noise data; otherwise, h t x in j is non-noise data; based on h t The non-noise data in constructs the training subset for the t-th training .

4. The intelligent health monitoring system for patient rehabilitation care according to claim 1 is characterized in that: The health intelligent monitoring model parameter optimization module is provided with an initial individual position unit, an individual position update unit and an optimal parameter determination unit, and specifically includes the following contents: Initial individual location unit, selection factor δ, coupling factor in health intelligent monitoring model , control factor ρ, update factor The parameter search space is established with the adjustment factor γ. The initial R individual positions are used as parameter representatives of the health intelligent monitoring model. The cross entropy loss of the health intelligent monitoring model based on parameter training for the test data set is used as the fitness value of the individual position. The chaotic map is combined with the sine function, and a random disturbance term that obeys the normal distribution is added. The formula used for the initial individual position is as follows: ； In the formula, and are the initial r+1th and rth individual positions respectively, r is the individual position index, sin(•) is the sine function, z is the chaos coefficient, It has a mean of 0 and a variance of σ 2 The random disturbance term of the normal distribution is is the disturbance coefficient; Update the individual position unit, randomly generate a judgment value for each individual position, and update the individual position based on the judgment value and the global optimal position. The formula used is as follows: ； In the formula, and are the rth individual positions in the p+1th and pth searches, respectively. p is the search count index, and v is the individual position index. and are the global optimal positions of the rth and vth individual positions since p searches, is the global optimal position among the R individual positions in the p-th search, is the judgment value of the rth individual position in the pth search, η, λ and are three random numbers that do not interfere with each other; Determine the optimal parameter unit, pre-set the fitness threshold and the maximum search times, update the fitness value of the individual position, and when the fitness value of the individual position X is lower than the fitness threshold, the model parameters represented by the individual position X are the optimal parameters, and a health intelligent monitoring model is constructed based on the optimal parameters; otherwise, if the maximum search times are reached, return to the initial individual position unit; otherwise, add 1 to the search times and return to the updated individual position unit to continue searching.

5. The intelligent health monitoring system for patient rehabilitation care according to claim 1 is characterized in that: The patient rehabilitation care module collects real-time patient medical data, which includes vital signs data, activity data, life data and pain level data. After data preprocessing, the real-time patient medical data is input into a health intelligent monitoring model constructed based on optimal parameters for classification to obtain the patient's health status. Medical staff adjust the rehabilitation care plan according to the patient's health status.

6. The intelligent health monitoring system for patient rehabilitation care according to claim 1, characterized in that: The medical data collection module collects historical patient medical data; the historical patient medical data includes vital signs data, activity data, life data, pain level data and health status; the health status is used as a data label.

7. The intelligent health monitoring system for patient rehabilitation care according to claim 1 is characterized in that: The medical data preprocessing module performs data cleaning, data conversion, data normalization and data set construction on the collected patient medical data; the data cleaning includes processing missing values, abnormal values ​​and duplicate values; the data conversion is to convert the data into vector form; the data normalization is to unify the data range based on the maximum and minimum normalization method; The constructed data set is to construct a training data set and a test data set based on the patient medical data after data cleaning, data conversion, and data normalization.

Images