Rehabilitation promotion effect evaluation system for postoperative nutrition nursing
By constructing a calibration skewness with clinical elasticity coefficients and supplementing with three-stage samples, combining the rehabilitation level distribution and bias design loss function, and using the XGBoost model, the benchmark bias and accuracy problems of the rehabilitation assessment system were solved, achieving a more reliable and accurate assessment of rehabilitation effects.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-30
- Publication Date
- 2026-04-10
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing rehabilitation promotion effect evaluation systems suffer from several drawbacks: the benchmark points for rehabilitation indicators are affected by extreme values, and the borderlines of indicators lack clinical limitations, leading to a deviation in the evaluation benchmark and low reliability of the effect evaluation; the system also has low discrimination for borderline cases and is affected by extreme deviation samples, resulting in poor accuracy of the effect evaluation.
A calibration skewness parameter with clinical elasticity coefficient was constructed, a three-stage sample supplementation module was divided, a rehabilitation level distribution was introduced, a rehabilitation deviation design grading loss function was defined, and the XGBoost model was used for evaluation. A regularization term was added to optimize the model parameters.
This improves the reliability and accuracy of rehabilitation promotion effect assessment, ensures that the sample is appropriate for the rehabilitation process, filters out interference from extreme cases, focuses on borderline patients, and avoids overfitting.
Smart Images
Figure CN121839015A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing, specifically to a system for evaluating the rehabilitation promotion effect of postoperative nutritional care. Background Technology
[0002] Rehabilitation effectiveness assessment systems typically collect data such as patients' physiological indicators and rehabilitation progress records, and use data analysis to evaluate the role of interventions such as nutritional support and rehabilitation training in promoting patient recovery. However, general rehabilitation effectiveness assessment systems suffer from several problems: the baseline of rehabilitation indicators is affected by extreme values, and the threshold of indicators lacks clinical definition, leading to baseline deviation and low reliability of effectiveness assessment; generally, these systems also have low discrimination against borderline cases and are affected by extreme deviation samples, resulting in poor accuracy of effectiveness assessment. Summary of the Invention
[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a rehabilitation promotion effect evaluation system for postoperative nutritional care. Addressing the problems of general rehabilitation promotion effect evaluation systems, such as benchmark points for rehabilitation indicators being affected by extreme values and the lack of clinically defined thresholds for indicators, leading to benchmark shifts and low reliability of effect evaluation, this solution constructs a calibration skewness parameter with a clinical elasticity coefficient. This allows the uneven distribution to accurately adapt to the rehabilitation pattern of the majority being fast and the minority being slow, generating samples that cover both routine rapid rehabilitation and slow rehabilitation with complications. It anchors to clinical reality based on defined thresholds, avoiding the generation of invalid samples. By dividing the system into three stages and anchoring the core objectives of each stage, it supplements the samples with stage-differentiated data. Furthermore, it incorporates the rehabilitation level distribution, ensuring that the generated samples in the observation area align with the rehabilitation process, while the samples in the distribution area follow the clinical pathway. To ensure that the generated samples are free of clinical inconsistencies and closely reflect the rehabilitation process during the transition phases, thereby improving the reliability of the final rehabilitation promotion effect assessment, this solution addresses the issue that general rehabilitation promotion effect assessment systems suffer from low discrimination against borderline cases and are susceptible to interference from extreme deviation samples, leading to poor accuracy. This solution defines a graded loss function based on rehabilitation deviation, focusing on borderline cases and weakening deviation samples. For borderline cases, a linear penalty is applied, with higher penalty intensity forcing focus on such samples and improving discrimination. For extreme deviation samples, a decreasing loss is used to reduce the penalty for deviation samples and avoid overfitting to extreme cases. The assessment focuses on borderline patients most sensitive to rehabilitation promotion while filtering out interference from extreme cases. A regularization term is added to the objective function to prevent overfitting, thus improving the accuracy of the rehabilitation promotion effect assessment.
[0004] The technical solution adopted by the present invention is as follows: The present invention provides a rehabilitation promotion effect evaluation system for postoperative nutritional care, including a data acquisition module, a critical calibration module, a phased sample supplementation module, a rehabilitation promotion effect evaluation model construction module, and a rehabilitation promotion effect evaluation module;
[0005] The data acquisition module obtains historical postoperative nutritional care data and constructs an original sample set.
[0006] The critical calibration module achieves critical calibration by introducing a calibration skew amount containing a clinical elasticity coefficient and a defined critical value.
[0007] The phased sample supplementation module is based on the division of observation area and distribution area. The observation area is judged by critical samples, and the distribution area is guided by clinical pathway to generate samples respectively, thus obtaining a rehabilitation assessment sample set.
[0008] The rehabilitation promotion effect evaluation model construction module is based on the rehabilitation evaluation sample set and XGBoost, and introduces the postoperative rehabilitation deviation grading loss to establish a rehabilitation promotion effect evaluation model.
[0009] The rehabilitation promotion effect evaluation module is based on the rehabilitation promotion effect evaluation model and evaluates the rehabilitation promotion effect on real-time postoperative nutritional care data.
[0010] Furthermore, the data acquisition module acquires historical postoperative nutritional care data, which includes nutritional status indicators, physiological indicators, length of hospital stay, and recovery level. The recovery level is used as a data label and standardized. The recovery level is processed using one-hot encoding to obtain the original sample set.
[0011] Furthermore, the critical calibration module performs a reference point calculation, then a calibration skew calculation, incorporates a clinical elasticity coefficient, and performs a limited critical calculation.
[0012] Furthermore, the phased sample supplementation module specifically includes:
[0013] Regional division involves dividing the original sample set into an observation area and a distribution area; for the samples in the observation area, let m be the number of similar cases at the same recovery stage, m + The number of delayed recovery samples among similar cases; the dispersion of the recovery level distribution is introduced into the number of similar cases; if If so, it is determined to be a critical sample;
[0014] Sample generation involves generating observation area samples based on critical samples; uniform distribution guided by clinical pathways is used for distribution area samples; labels corresponding to synthesized samples are inherited from source sample labels; and the collected original samples, supplementary observation area samples, and distribution area samples are merged into a rehabilitation assessment sample set.
[0015] Furthermore, the rehabilitation promotion effect evaluation model construction module is based on the rehabilitation assessment sample set and XGBoost, introducing postoperative rehabilitation deviation grading loss, and thus establishing a rehabilitation promotion effect evaluation model; specifically including:
[0016] Deviation quantification indicators, defining deviation;
[0017] Loss function design, based on deviation, for postoperative rehabilitation deviation grading loss design;
[0018] Optimize the objective function and calibrate it using the postoperative rehabilitation deviation grading loss.
[0019] Model training involves optimizing model parameters using Newton's method; node splitting gain is calculated based on loss gradient and second derivative, and the feature with the largest gain and threshold are selected for tree splitting; model performance is evaluated by classification accuracy. If the target is met, the rehabilitation promotion effect evaluation model is established; otherwise, the initial parameters are optimized using particle swarm optimization.
[0020] Furthermore, the rehabilitation promotion effect assessment module acquires postoperative nutritional care data in real time, excluding rehabilitation level, and inputs it into the rehabilitation promotion effect assessment model after preprocessing. The model outputs the rehabilitation promotion effect assessment results.
[0021] The beneficial effects achieved by the present invention using the above solution are as follows:
[0022] (1) To address the problem that general rehabilitation promotion effect evaluation systems suffer from the interference of extreme values in the benchmark points of rehabilitation indicators and the lack of clinical limitations on the threshold of indicators, which leads to the deviation of the evaluation benchmark and low reliability of the effect evaluation, this solution constructs a calibration skewness with a clinical elasticity coefficient to allow the uneven distribution to accurately adapt to the rehabilitation pattern of the majority being fast and the minority being slow. The generated samples can cover two scenarios: routine rapid rehabilitation and slow rehabilitation with complications. Based on the limited threshold, it anchors to clinical reality to avoid generating invalid samples. By dividing into three stages, it anchors the core goals of each stage and supplements the stage-differentiated samples. It introduces the combination of rehabilitation level distribution, the sample generation in the observation area is in line with the rehabilitation process, and the sample generation in the distribution area follows the clinical pathway to ensure that the generated samples have no clinical contradictions and are in line with the rehabilitation process of the stage transition. This improves the reliability of the final rehabilitation promotion effect evaluation.
[0023] (2) To address the problem that general rehabilitation promotion effect evaluation systems have low discrimination for borderline cases and are affected by extreme deviation samples, resulting in poor accuracy of effect evaluation, this solution defines a graded loss function for rehabilitation deviation design, focuses on borderline cases and weakens deviation samples. For borderline cases, a linear penalty is applied, with a relatively higher penalty intensity to force focus on such samples and improve discrimination. For extreme deviation samples, a decreasing loss is applied to reduce the penalty for deviation samples and avoid overfitting extreme cases. The evaluation focuses on borderline patients who are most sensitive to rehabilitation promotion, while filtering out interference from extreme cases. A regularization term is added to the objective function to avoid overfitting, thereby improving the accuracy of rehabilitation promotion effect evaluation. Attached Figure Description
[0024] Figure 1 This invention provides a flowchart illustrating a system for evaluating the rehabilitation promotion effect of postoperative nutritional care.
[0025] Figure 2 A flowchart illustrating the module for constructing a rehabilitation promotion effect evaluation model.
[0026] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation
[0027] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0028] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.
[0029] Example 1, see Figure 1 The present invention provides a rehabilitation promotion effect evaluation system for postoperative nutritional care, including a data acquisition module, a critical calibration module, a phased sample supplementation module, a rehabilitation promotion effect evaluation model construction module, and a rehabilitation promotion effect evaluation module;
[0030] The data acquisition module obtains historical postoperative nutritional care data, constructs an original sample set, and sends the data to the critical calibration module.
[0031] The critical calibration module achieves critical calibration by introducing a calibration skew amount containing a clinical elasticity coefficient and a defined critical value; and sends the data to the phased sample replenishment module.
[0032] The phased sample supplementation module is based on the division of observation area and distribution area. The observation area is judged by critical samples, and the distribution area is guided by clinical pathway to generate samples respectively, so as to obtain a rehabilitation assessment sample set; and send the data to the rehabilitation promotion effect assessment model construction module.
[0033] The rehabilitation promotion effect evaluation model construction module is based on the rehabilitation evaluation sample set and XGBoost, introduces the postoperative rehabilitation deviation grading loss, establishes a rehabilitation promotion effect evaluation model, and sends the data to the rehabilitation promotion effect evaluation module.
[0034] The rehabilitation promotion effect evaluation module is based on the rehabilitation promotion effect evaluation model and evaluates the rehabilitation promotion effect on real-time postoperative nutritional care data.
[0035] Example 2, see Figure 1 This embodiment is based on the above embodiment. The data acquisition module acquires historical postoperative nutritional care data. The historical postoperative nutritional care data includes nutritional status indicators, physiological indicators, length of hospital stay, and recovery level. The nutritional status indicators include serum albumin, prealbumin, hemoglobin, body mass index, upper arm circumference, and triceps skinfold thickness. The physiological indicators include body temperature, heart rate, systolic / diastolic blood pressure, respiratory rate, fasting blood glucose, C-reactive protein, and white blood cell count. The recovery level is used as a data label, and the recovery level includes excellent, good, moderate, and poor. The data is standardized using Z-score standardization. The recovery level is then processed using one-hot encoding to obtain the original sample set.
[0036] Example 3, see Figure 1 This embodiment is based on the above embodiment. The critical calibration module addresses the uneven distribution of postoperative rehabilitation indicators, including length of hospital stay and albumin levels. Most patients are discharged within 10 days, but a few require more than 30 days. Traditional symmetrical distribution can lead to deviations in the generated nursing samples from clinical reality. Therefore, critical calibration is performed based on the precise uneven distribution of the rehabilitation indicator range. Specifically, the baseline point is calculated using the median, which has more clinical significance, instead of the mean, to avoid interference from extreme values in long-term hospitalized patients. The baseline point JZ is represented as: ; This is the original sample set sorted in ascending order by length of hospital stay, where n is the total number of samples in the original sample set; the calibration skewness is calculated by incorporating the clinical elasticity coefficient. (Values range from 0.1 to 1.0), to prevent overestimation of deviation caused by postoperative complications, as expressed as: ; ;in, It is the number of samples below the baseline. It is the number of samples higher than the baseline. It is a specific calibration factor, with a value ranging from 0.01 to 0.2; and These are the calibration skew amounts below and above the reference point, respectively; and a limiting critical calculation is performed, including the lower critical value. Represented as: Corresponding clinical warning line; upper threshold Represented as: ;in, It is the sample variance; It is the initial lower critical point; It is the clinical warning line; It is the initial upper critical point; This is the upper limit of clinical practice; By constraining the clinical boundaries of critical values with extremely low probabilities, and adapting the uneven distribution to the actual distribution of postoperative recovery where most recover quickly and a few recover slowly, the benchmark calculation is not affected by extreme values, making it more suitable for medical data of cases with complications.
[0037] Example 4, see Figure 1 This embodiment is based on the above embodiment. The phased sample supplementation module divides the postoperative recovery stage into three phases: acute phase (1-7 days), subacute phase (8-30 days), and recovery phase (31 days and above). The nutrition-rehabilitation correlation pattern is different in each phase, therefore the divided observation area and distribution area are supplemented differentially; specifically including:
[0038] Regional division involves dividing the original sample set into observation areas and distribution areas; the observation areas correspond to... ; and These are the lower and upper limits of the observation area, respectively; the corresponding distribution area... and For the sample in the observation area, let m be the number of similar cases in the same recovery stage, m + The number of samples with delayed recovery among similar cases; the number of similar cases introduces the dispersion of the recovery level distribution. The greater the dispersion, the more dispersed the recovery levels, and the more similar cases are needed to determine the critical level, expressed as: ; ;in, It is the number of basic similar cases; It represents the total number of samples in recovery phase d. It is the influence coefficient; It is the dispersion of the rehabilitation level distribution; This represents the percentage of samples with a rehabilitation level of y within rehabilitation stage d; the delayed rehabilitation sample refers to samples where, within the same rehabilitation stage, core rehabilitation indicators deviate from the direction conducive to rehabilitation. For positive indicators such as serum albumin, prealbumin, and hemoglobin, these indicators are below 90% of the average for the same stage; for negative indicators such as C-reactive protein and fasting blood glucose, these indicators are above 110% of the average for the same stage. If it is, it is determined to be a critical sample, which includes fuzzy samples that transition from the acute phase to the subacute phase.
[0039] Sample generation, represented as: ; It is a critical sample; These are samples generated in the observation area; These are similar case samples from the same stage; It generates random numbers; This is a phase coefficient, with a value of 0.6 for the acute phase, 0.4 for the subacute phase, and 0.2 for the recovery phase. The sample generation for the distribution area uses a uniform distribution guided by clinical pathways. Samples are generated for low-sample areas according to clinical rehabilitation pathways, including nutritional recovery data from patients who have been fasting for a long period post-surgery, and are represented as follows: ; ; ; These are samples generated from the distribution area; a and b are distribution area coefficients; for the labels corresponding to the synthesized samples, the source sample labels are inherited, and if the parent labels are different, the nearest parent label is inherited first; the collected original samples, supplementary observation area samples, and distribution area samples are merged into a rehabilitation assessment sample set;
[0040] Samples are generated based on the stage-specific distinction of rehabilitation stages to avoid confusion about the differences in nutritional needs at different stages; the judgment of critical samples is combined with medical characteristics, and the generated samples are more consistent with the actual rehabilitation process.
[0041] By performing the above operations, this solution addresses the problems of general rehabilitation effectiveness assessment systems, such as the benchmark points of rehabilitation indicators being affected by extreme values and the lack of clinical limitations on indicator thresholds, leading to benchmark deviations and low reliability of effectiveness assessments. It constructs a calibration skewness parameter with a clinical elasticity coefficient, allowing the uneven distribution to accurately adapt to the rehabilitation pattern of the majority being fast and the minority slow. The generated samples can cover both routine rapid rehabilitation and slow rehabilitation with complications. It anchors to clinical reality based on defined thresholds, avoiding the generation of invalid samples. By dividing the system into three stages and anchoring the core objectives of each stage, it supplements the samples with stage-differentiated data. Furthermore, it incorporates the distribution of rehabilitation levels, ensuring that samples in the observation area align with the rehabilitation process and samples in the distribution area follow clinical pathways, guaranteeing that the generated samples are free of clinical contradictions and accurately reflect the rehabilitation process during stage transitions. This ultimately improves the reliability of the final rehabilitation effectiveness assessment.
[0042] Example 5, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. The rehabilitation promotion effect evaluation model construction module is based on the rehabilitation evaluation sample set and XGBoost, and introduces postoperative rehabilitation deviation grading loss to achieve high weight for samples close to the rehabilitation decision threshold and low weight for deviation samples far from the threshold; thereby establishing a rehabilitation promotion effect evaluation model; specifically including:
[0043] Deviation quantification indicators; defining deviation ; ; This reflects the actual level of recovery. is the predicted recovery level for the i-th patient in round t; d is the recovery level distance, which refers to the absolute difference between the actual recovery level and the predicted recovery level. It is a rehabilitation level matching function;
[0044] Loss function design; postoperative rehabilitation deviation grading loss design, quantifying the deviation between the rehabilitation characteristics and the expected outcome label of an individual patient, postoperative rehabilitation deviation grading loss. Represented as: ; g is the adaptation parameter, taking values from 1 to 5; when Losses follow Increasing while decreasing, the penalty for deviations far from the critical point is reduced; when Focusing on borderline rehabilitation cases, linear punishment enhances differentiation;
[0045] Optimize the objective function; calibrate the objective function using the postoperative rehabilitation deviation grading loss, optimize it through gradient boosting, and enhance the nonlinear fitting ability using kernel features. The objective function in the t-th round is expressed as: ; ;in, This is the total loss function for round t; n is the total number of samples, and i is the sample index; It is the bias of the i-th sample in the (t-1)-th iteration; It is a regularization term. It is the newly added decision tree in round t; is the penalty coefficient, ranging from 0 to 1; T is the number of leaf nodes in the t-th decision tree; This is the regularization coefficient, with a value ranging from 0 to 10; It is the output weight of the j-th leaf node;
[0046] Model training; model parameters are optimized using Newton's method; node splitting gain is calculated based on loss gradient and second derivative, and the feature with the largest gain and threshold are selected for tree splitting, as shown below: Where Gn is the gain value, used to balance node splitting; It is the first derivative of the loss of the i-th sample in the t-th round. is the second derivative; L and R are the sample sets of the left and right subtrees after the node split, respectively; all is the sample set of the parent node before the node split; the model performance is evaluated by the classification accuracy. If the target is met (threshold value is 0.8~0.95), the rehabilitation promotion effect evaluation model is established. Otherwise, the initial parameters are optimized by the particle swarm optimization algorithm.
[0047] By performing the above operations, this solution addresses the problems of low discrimination against borderline cases and interference from extreme deviation samples in general rehabilitation promotion effect assessment systems, leading to poor accuracy in effect assessment. It defines a graded loss function based on rehabilitation deviation, focusing on borderline cases and weakening deviation samples. For borderline cases, a linear penalty is applied, with higher penalty intensity forcing focus on such samples and improving discrimination. For extreme deviation samples, a decreasing loss is used to reduce the penalty for deviation samples and avoid overfitting to extreme cases. The assessment focuses on borderline patients most sensitive to rehabilitation promotion while filtering out interference from extreme cases. A regularization term is added to the objective function to avoid overfitting, thereby improving the accuracy of rehabilitation promotion effect assessment.
[0048] Example 6, see Figure 1 This embodiment is based on the above embodiment. The rehabilitation promotion effect evaluation module acquires postoperative nutritional care data in real time, excluding rehabilitation level. After preprocessing, the data is input into the rehabilitation promotion effect evaluation model. The model outputs the rehabilitation promotion effect evaluation results. If the model output is medium, the sampling frequency is increased. If the model output is poor, an early warning is issued to the relevant personnel.
[0049] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.
[0050] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.
Claims
1. A rehabilitation promotion effect evaluation system for postoperative nutritional care, characterized by: The system comprises a data acquisition module, a critical calibration module, a staged sample supplement module, a rehabilitation promotion effect evaluation model construction module and a rehabilitation promotion effect evaluation module. The data acquisition module acquires historical postoperative nutritional care data and constructs an original sample set. The critical calibration module realizes critical calibration by introducing a calibration bias containing a clinical elasticity coefficient and limiting a critical value. The staged sample supplement module generates samples based on the observation area and the distribution area, judges the observation area through critical samples, and guides the distribution area using a clinical pathway to obtain a rehabilitation evaluation sample set. The rehabilitation promotion effect evaluation model construction module introduces a postoperative rehabilitation bias classification loss based on the rehabilitation evaluation sample set and XGBoost, and establishes a rehabilitation promotion effect evaluation model. The rehabilitation promotion effect evaluation module evaluates the rehabilitation promotion effect of real-time postoperative nutritional care data based on the rehabilitation promotion effect evaluation model.
2. The rehabilitation promotion effect evaluation system for postoperative nutritional care according to claim 1, characterized by: The critical calibration module calculates the reference point, then calculates the calibration bias, adds the clinical elasticity coefficient, and limits the critical value.
3. The rehabilitation promotion effect evaluation system for postoperative nutritional care according to claim 2, characterized by: The staged sample supplement module specifically comprises: Region division, divide observation area and distribution area for original sample set; for observation area sample, set m as the number of similar cases in rehabilitation stage, m + is the number of rehabilitation delay samples in similar cases; the number of similar cases introduces the dispersion of rehabilitation level distribution; if , it is determined as critical sample; Sample generation, observation area sample generation based on critical samples; for distribution area samples, uniform distribution guided by a clinical pathway; for the corresponding labels of the synthesized samples, follow the inheritance of the source sample labels; and the acquired original samples, supplemented observation area samples and distribution area samples are merged into a rehabilitation evaluation sample set.
4. The rehabilitation promotion effect evaluation system for postoperative nutritional care according to claim 3, characterized by: The rehabilitation promotion effect evaluation model construction module introduces a postoperative rehabilitation bias classification loss based on the rehabilitation evaluation sample set and XGBoost, and further establishes a rehabilitation promotion effect evaluation model; specifically comprising: Bias quantification index, define bias; Loss function design, design postoperative rehabilitation bias classification loss based on bias; Optimization objective function, calibrate the objective function with the postoperative rehabilitation bias classification loss; Model training, model parameters are optimized using the Newton method; based on the loss gradient and the second-order derivative calculation node split gain, select the feature with the maximum gain and the threshold for tree splitting; evaluate the model performance by classification accuracy, if the standard is met, the rehabilitation promotion effect evaluation model is established, otherwise, optimize the initial parameters through the particle swarm algorithm.
5. The rehabilitation promotion effect evaluation system for postoperative nutritional care according to claim 4, characterized by: The data acquisition module acquires historical postoperative nutritional care data; the historical postoperative nutritional care data includes nutritional status indicators, physiological indicators, hospitalization days and rehabilitation levels; the rehabilitation level is used as a data label; And perform standardization processing; use one-hot encoding processing for the rehabilitation level; obtain the original sample set.
6. The rehabilitation promotion effect evaluation system for postoperative nutritional care according to claim 5, characterized by: The rehabilitation promotion effect evaluation module is to acquire real-time postoperative nutritional care data, excluding the rehabilitation level, and input the preprocessed data into the rehabilitation promotion effect evaluation model to obtain the rehabilitation promotion effect evaluation result.