Nursing data automated collection and analysis method and system

Abstract

The present invention discloses a method and system for automatic collection and analysis of nursing data, and relates to the technical field of data collection and analysis. The automatic collection and analysis system for nursing data includes: a data collection module, a data storage module, a patient data mining module, a mining optimization module, an evaluation and prediction module, and a dynamic scheduling module. The present invention effectively reduces the storage of redundant data through the compression storage technology of nursing data, especially performs hidden data processing on data with small changes, significantly saves storage space, reduces storage costs, and improves data storage efficiency; uses intelligent data mining algorithms to analyze patient nursing data in real time, extract key health indicators and change trends, provide nursing staff with accurate patient status assessment and disease prediction support, and improve data utilization and timeliness of decision-making.

Description

Technical Field

[0001] The present invention relates to the field of data collection and analysis, and in particular to a method and system for automatic collection and analysis of nursing data. Background Art

[0002] With the development of medical information technology, a large amount of data is generated in nursing work, including nurses' work logs, patients' physiological indicators, etc. The management and analysis of these data play a vital role in nursing work. However, the existing nursing data management methods mainly rely on manual operations, data storage is redundant, analysis efficiency is low, and the value of data cannot be fully utilized. In addition, a large amount of storage resources are occupied, affecting the overall performance of the system. After the collection of nursing data, how to store the data efficiently is a major problem. The traditional data storage method directly saves all data, resulting in serious waste of storage space. In addition, nursing data contains a large amount of patient health information. How to use this data for effective mining and discover potential health risks and trends is another problem in the existing technology. When the existing nursing data mining model processes complex nursing data, the model is prone to fall into the local optimal solution, which affects the accuracy of the mining results. At the same time, the scheduling management in nursing work also faces challenges. The existing scheduling method lacks flexibility and cannot be dynamically adjusted according to the real-time nursing needs of patients.

[0003] Therefore, developing a system that can automatically collect, preprocess, store and analyze nursing data in order to provide real-time and intelligent support for nursing decisions is of great significance for improving the efficiency and quality of nursing services. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed.

[0005] The present application provides a method for automatic collection and analysis of nursing data, which is characterized by comprising the following steps:

[0006] Step S1: Automatically collect nursing data, including nurse data and patient data. Nurse data is obtained through the nursing staff's work log, including the nursing staff's identity and level, working hours, shift mark and leave data; patient data is automatically recorded by various sensors equipped in the nursing equipment and filled in by the nursing staff. The automatically recorded patient data includes the patient's body temperature, blood pressure, heart rate, blood oxygen saturation and disease corresponding data. The filled in patient data includes blood sugar measurement value, blood lipid measurement value and disease corresponding data; the acquired patient data is the first patient data;

[0007] Step S2: preprocess the acquired first patient data, and use the preprocessed first patient data as the second patient data; calculate the change rate for each item of the second patient data, and divide the second patient data into periodic numerical data, non-periodic numerical data and unstructured data according to the data type. The unstructured data is marked with an evaluation category label. For the non-periodic numerical data, the difference percentage of adjacent data is calculated as the change rate. For the periodic numerical data, the root mean square error of adjacent periods is calculated, and the difference percentage between the root mean square error and the mean of the periodic data is used as the change rate. For the unstructured data, the change rate is obtained by changing the category label of the adjacent data, and the change rate is recombined with the second patient data to obtain the third patient data. According to the third patient data, a real-time evaluation of the patient data is performed to obtain the real-time evaluation result of the patient data; the basic pathological information of the patient is obtained, and the third patient data is stored and detected. The storage detection includes a first storage detection and a second storage detection, and data change detection of data points and data segments is performed respectively; the third patient data that passes the storage detection is compressed and stored, and the stored data is used as the fourth patient data. The data in the third patient data that is not completely stored in the fourth patient data is called hidden data;

[0008] Step S3: organizing the fourth patient data collected historically into a nursing data mining data set, performing data change trend analysis on the nursing data mining data set, adding the trend analysis results to the latent data in the nursing data mining data set, using the trend analysis results as data features of the latent data, and using the nursing data mining data set to train an initial nursing data mining model;

[0009] Step S4: Optimize the category attribute association of the initial nursing data mining model, obtain the optimal pairing parameters of each classification category and the data set of the initial nursing data mining model, and use the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm to optimize the hyperparameters of the initial nursing data mining model. The algorithm adopts a two-stage diversity adjustment mechanism in the search process to avoid the model parameters from falling into the local optimal solution, and obtain the nursing data mining data model;

[0010] Step S5: sending the patient data acquired in real time to the nursing data mining data model for calculation to obtain the nursing data analysis results, and evaluating the patient status and predicting the disease according to the nursing data analysis results;

[0011] Step S6: Obtain the patient's estimated nursing time based on the patient's status assessment and disease prediction results, and dynamically schedule the nurses based on the total estimated nursing time of all patients and nurse data.

[0012] As a preferred technical solution of the present invention, the storage detection includes:

[0013] The average value of the data with a change rate less than 15% in each data of the patient is obtained through the third patient data as the basic status data of the patient, the change rate detection threshold of each data is obtained through the basic status data of the patient, the threshold adjustment weight of the corresponding data is obtained according to the basic pathological information of the patient, and the change rate detection threshold is weightedly calculated to obtain the personalized detection threshold of the patient;

[0014] a first storage detection, comparing a rate of change of data in the third patient data with a personalized detection threshold, and first marking data exceeding the personalized detection threshold;

[0015] Second storage detection, using a window of size N to slide and select each item of data in the third patient data, accumulating the change rate of the data in the window, and when the total change rate of the data in the window exceeds N / K*personalized detection threshold, where K is the expected control value of the storage change rate, the first data and the last data in the window are second marked;

[0016] The data of the first mark and the second mark are compressed and stored.

[0017] As a preferred technical solution of the present invention, the compressed storage includes:

[0018] The third patient data is stored in a table format, where each column of the table represents each item of the third patient data. For each item of the third patient data, the data at the first moment is stored, and the data with the first mark and the second mark are stored. The data not stored in the middle is regarded as hidden data. The hidden data is represented by adding a quantity code after the stored data, and its content is consistent with the corresponding stored data. All items of data are merged into a table, and the table is converted into a coded format for storage.

[0019] As a preferred technical solution of the present invention, the initial nursing data mining model includes:

[0020] The initial nursing data mining model is constructed based on long-short time neural networks. The model includes an input layer for inputting data, L hidden layers for processing data connections, and 1 output layer for outputting classification results. The number of hidden layers in the model is determined by the number of data items in the patient data. The specific number of hidden layers is the number of combinations of three random elements of all data items, the number of classification hidden layers of specific categories, and 1 hidden layer of the final classification result. After the initial model structure is constructed, the model is trained.

[0021] As a preferred technical solution of the present invention, the training of the initial nursing data mining model includes:

[0022] The nursing data mining data set is obtained by decoding and restoring the stored fourth patient data, the change trend of the data is obtained according to the adjacent labeled data in the data set, the change trend is reorganized with the latent data, and an analysis label is added to the data corresponding to each patient, and the nursing data mining data set is divided into a training set, a validation set and a test set;

[0023] The initial nursing data mining model is optimized for category attribute association, and the nursing data mining data set is sent to a pre-trained classification model for calculation to obtain the classification results of the data set, and the weights of each data item corresponding to the classification results in the hidden layer of the model are extracted. The samples corresponding to the same analysis label are statistically analyzed for data weights, and the corresponding data items with the top three weight statistics are obtained as the optimal pairing parameters between the classification category and the data set; if the difference between the top three data in the weight statistics result is too large, the corresponding analysis label is used as a specific category; the analysis parameters are assigned to the hidden layer of the initial nursing data mining model according to the optimal pairing parameters and the specific category;

[0024] The nursing data mining training set and the validation set are sent to the initial nursing data mining model for training, with the analysis of labels as the goal, and the optimized nursing data mining model is obtained through training;

[0025] The nursing data mining test set is sent to the optimized nursing data mining model for accuracy evaluation. If the evaluation result exceeds the set threshold, the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm is used to optimize the hyperparameters of the initial nursing data mining model until the evaluation result is within the set range to obtain the final nursing data mining model.

[0026] As a preferred technical solution of the present invention, the optimization of hyperparameters includes:

[0027] Construct a set of uniformly distributed weight vectors, and use the FBI method to decompose the target optimization problem into a series of single-target sub-problems; randomly generate the initial position vector of the particle, obtain the initial individual best position of the particle, calculate the average best position of the population, and calculate the global optimal position based on the Pareto dominance relationship between the historical individual best positions of different particles;

[0028] Define a set GS, whose size is n(GS), and use the fast non-dominated sorting method to sort the elements in the set GS and all the current individual best position vectors. The higher the non-dominated ranking, the better the quality of the vector. The individual best position vector with the highest ranking is selected and stored in the set GS. All vectors in the set GS are regarded as candidate global optimal solutions and are randomly applied to the next iteration to guide the search of each particle in the population in the target space to obtain the global optimal position.

[0029] As a preferred technical solution of the present invention, dynamic scheduling includes:

[0030] After obtaining the nursing data analysis results of the patient, match the patient status and disease prediction results in the personalized status assessment strategy library through the patient's personal information, calculate the patient's estimated nursing time based on the matched patient status and disease prediction results, and calculate the total estimated nursing time for all patients;

[0031] Obtain nurse data and dynamically schedule according to each nurse's monthly working hours, weekly working hours, shift tags and levels.

[0032] As a preferred technical solution of the present invention, unstructured data includes: image data, text data, and sound data. The evaluation category labels of unstructured data are divided into major categories and levels, and their change rates are scalar calculated through the differences between major categories and levels.

[0033] A nursing data automatic collection and analysis system, including the following modules:

[0034] The data collection module is used to automatically collect nursing data, including nurse data and patient data. The nurse data is obtained through the nursing staff's work log, including the nursing staff's identity and level, working hours, shift mark and leave data; the patient data is automatically recorded by various sensors equipped in the nursing equipment and filled in by the nursing staff. The automatically recorded patient data includes the patient's body temperature, blood pressure, heart rate, blood oxygen saturation and disease corresponding data. The filled in patient data includes blood sugar measurement value, blood lipid measurement value and disease corresponding data; the acquired patient data is the first patient data;

[0035] The data storage module includes a data preprocessing unit, a change rate calculation unit, a real-time evaluation unit and a storage detection unit. The data preprocessing unit is used to perform data preprocessing on the acquired first patient data, and the preprocessed first patient data is used as the second patient data; the change rate calculation unit is used to calculate the change rate of each data item of the second patient data, and the second patient data is divided into periodic numerical data, non-periodic numerical data and unstructured data according to the data type. The unstructured data is marked with an evaluation category label. For the non-periodic numerical data, the difference percentage of adjacent data is calculated as the change rate. For the periodic numerical data, the root mean square error of adjacent periods is calculated, and the difference percentage between the root mean square error and the mean of the periodic data is calculated. The ratio of the data points is used as the change rate. For unstructured data, the change rate is obtained by changing the category labels of adjacent data. The change rate is recombined with the second patient data to obtain the third patient data. The real-time evaluation unit is used to perform real-time evaluation of the patient data according to the third patient data to obtain the real-time evaluation result of the patient data; the basic pathological information of the patient is obtained. The storage detection unit is used to perform storage detection on the third patient data. The storage detection includes a first storage detection and a second storage detection, which respectively perform data change detection on data points and data segments; the third patient data that passes the storage detection is compressed and stored, and the stored data is used as the fourth patient data. The data in the third patient data that is not completely stored in the fourth patient data is called hidden data;

[0036] A patient data mining module is used to organize the fourth patient data collected historically into a nursing data mining data set, perform data change trend analysis on the nursing data mining data set, add the trend analysis results to the latent data in the nursing data mining data set, use the trend analysis results as data features of the latent data, and use the nursing data mining data set to train an initial nursing data mining model;

[0037] The mining optimization module is used to optimize the category attribute association of the initial nursing data mining model, obtain the optimal pairing parameters of each classification category and the data set of the initial nursing data mining model, and use the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm to optimize the hyperparameters of the initial nursing data mining model. The algorithm adopts a two-stage diversity adjustment mechanism in the search process to avoid the model parameters from falling into the local optimal solution, and obtain the nursing data mining data model;

[0038] The evaluation and prediction module is used to send the real-time acquired patient data into the nursing data mining data model for calculation, obtain the nursing data analysis results, and evaluate the patient status and predict the disease based on the nursing data analysis results;

[0039] The dynamic scheduling module is used to obtain the patient's estimated nursing time based on the patient's status assessment and disease prediction results, and to perform dynamic scheduling based on the total estimated nursing time of all patients and nurse data.

[0040] The present invention has the following advantages:

[0041] The present invention effectively reduces the storage of redundant data through the compression storage technology of nursing data, especially performs hidden data processing on data with small changes, significantly saves storage space, reduces storage costs, and improves data storage efficiency; utilizes intelligent data mining algorithms to analyze patient nursing data in real time, extract key health indicators and changing trends, provide nursing staff with accurate patient status assessment and disease prediction support, and improve data utilization and timeliness of decision-making.

[0042] The present invention optimizes the hyperparameters of the data mining model based on a multi-objective quantum behavior particle swarm optimization algorithm, effectively improving the accuracy and stability of the model, preventing the model from falling into local optimality, ensuring the reliability of the mining results, and enhancing the in-depth analysis capabilities of nursing data.

[0043] The present invention performs intelligent calculations on nursing data analysis results, dynamically adjusts nurse scheduling, optimizes the allocation of nursing resources, and can flexibly schedule according to the real-time needs of patients and the working status of nurses, thereby improving the efficiency of nursing work and service quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the structure of a nursing data automatic collection and analysis system used in an embodiment of the present invention;

[0045] Figure 2 This is a structural diagram of a data storage module of a nursing data automatic collection and analysis system in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0047] Embodiment 1 of the present invention is a method for automatically collecting and analyzing nursing data, comprising the following steps:

[0048] Step S1: Automatically collect nursing data, including nurse data and patient data. Nurse data is obtained through the nursing staff's work log, including the nursing staff's identity and level, working hours, shift mark and leave data; patient data is automatically recorded by various sensors equipped in the nursing equipment and filled in by the nursing staff. The automatically recorded patient data includes the patient's body temperature, blood pressure, heart rate, blood oxygen saturation and disease corresponding data. The filled in patient data includes blood sugar measurement value, blood lipid measurement value and disease corresponding data; the acquired patient data is the first patient data;

[0049] The first type of disease-corresponding data is the patient data automatically recorded by the automatic data collection device in addition to the common data items (body temperature, blood pressure), and the second type of disease-corresponding data is the patient data recorded by manual measurement in addition to the common data (blood sugar, blood lipids). The two types of data obtain the data items to be collected according to the specific disease type of the patient;

[0050] Step S2: preprocess the acquired first patient data, and use the preprocessed first patient data as the second patient data; calculate the change rate for each item of the second patient data, and divide the second patient data into periodic numerical data, non-periodic numerical data and unstructured data according to the data type. The unstructured data is marked with an evaluation category label. For the non-periodic numerical data, the difference percentage of adjacent data is calculated as the change rate. For the periodic numerical data, the root mean square error of adjacent periods is calculated, and the difference percentage between the root mean square error and the mean of the periodic data is used as the change rate. For the unstructured data, the change rate is obtained by changing the category label of the adjacent data, and the change rate is recombined with the second patient data to obtain the third patient data. According to the third patient data, a real-time evaluation of the patient data is performed to obtain the real-time evaluation result of the patient data; the basic pathological information of the patient is obtained, and the third patient data is stored and detected. The storage detection includes a first storage detection and a second storage detection, and data change detection of data points and data segments is performed respectively; the third patient data that passes the storage detection is compressed and stored, and the stored data is used as the fourth patient data. The data in the third patient data that is not completely stored in the fourth patient data is called hidden data;

[0051] Unstructured data includes: image data, text data, and sound data. The evaluation category labels of unstructured data are divided into major categories and levels, and their change rates are scalar calculated through the differences between major categories and levels.

[0052] Storage testing includes:

[0053] The average value of the data with a change rate less than 15% in each data of the patient is obtained through the third patient data as the basic status data of the patient, the change rate detection threshold of each data is obtained through the basic status data of the patient, the threshold adjustment weight of the corresponding data is obtained according to the basic pathological information of the patient, and the change rate detection threshold is weightedly calculated to obtain the personalized detection threshold of the patient;

[0054] a first storage detection, comparing a rate of change of data in the third patient data with a personalized detection threshold, and first marking data exceeding the personalized detection threshold;

[0055] Second storage detection, using a window of size N to slide and select each item of data in the third patient data, accumulating the change rate of the data in the window, and when the total change rate of the data in the window exceeds N / K*personalized detection threshold, where K is the expected control value of the storage change rate, the first data and the last data in the window are second marked;

[0056] The data of the first mark and the second mark are compressed and stored.

[0057] Compressed storage includes:

[0058] The third patient data is stored in a table format, where each column of the table represents each item of the third patient data. For each item of the third patient data, the data at the first moment is stored, and the data with the first mark and the second mark are stored. The data not stored in the middle is regarded as hidden data. The hidden data is represented by adding a quantity code after the stored data, and its content is consistent with the corresponding stored data. All items of data are merged into a table, and the table is converted into a coded format for storage.

[0059] The quantity encoding is the number of corresponding hidden data. For example, a column of data is stored as follows: the first data T, the labeled data A-3, the labeled data B, and the labeled data C. The -3 in A-3 represents three hidden data that are consistent with the labeled data A. After restoration, the column of data is [T, A, A, A, A, B, C].

[0060] Step S3: organizing the fourth patient data collected historically into a nursing data mining data set, performing data change trend analysis on the nursing data mining data set, adding the trend analysis results to the latent data in the nursing data mining data set, using the trend analysis results as data features of the latent data, and using the nursing data mining data set to train an initial nursing data mining model;

[0061] The nursing data mining data set is obtained by decoding and restoring the fourth patient data, in which the hidden data part is still marked as hidden data after restoration, and the trend analysis results are added, specifically the increase and decrease step marks, which are obtained through the changes of the two marked data before and after the hidden data;

[0062] Step S4: Optimize the category attribute association of the initial nursing data mining model, obtain the optimal pairing parameters of each classification category and the data set of the initial nursing data mining model, and use the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm to optimize the hyperparameters of the initial nursing data mining model. The algorithm adopts a two-stage diversity adjustment mechanism in the search process to avoid the model parameters from falling into the local optimal solution, and obtain the nursing data mining data model;

[0063] The initial nursing data mining model included:

[0064] The initial nursing data mining model is constructed based on long-short time neural networks. The model includes an input layer for inputting data, L hidden layers for processing data connections, and 1 output layer for outputting classification results. The number of hidden layers in the model is determined by the number of data items in the patient data. The specific number of hidden layers is the number of combinations of three random elements of all data items, the number of classification hidden layers of specific categories, and 1 hidden layer of the final classification result. After the initial model structure is constructed, the model is trained.

[0065] The training of the initial nursing data mining model includes:

[0066] The nursing data mining data set is obtained by decoding and restoring the stored fourth patient data, the change trend of the data is obtained according to the adjacent labeled data in the data set, the change trend is reorganized with the latent data, and an analysis label is added to the data corresponding to each patient, and the nursing data mining data set is divided into a training set, a validation set and a test set;

[0067] The initial nursing data mining model is optimized for category attribute association, and the nursing data mining data set is sent to a pre-trained classification model for calculation to obtain the classification results of the data set, and the weights of each data item corresponding to the classification results in the hidden layer of the model are extracted. The samples corresponding to the same analysis label are statistically analyzed for data weights, and the corresponding data items with the top three weight statistics are obtained as the optimal pairing parameters between the classification category and the data set; if the difference between the top three data in the weight statistics result is too large, the corresponding analysis label is used as a specific category; the analysis parameters are assigned to the hidden layer of the initial nursing data mining model according to the optimal pairing parameters and the specific category;

[0068] The nursing data mining training set and the validation set are sent to the initial nursing data mining model for training, with the analysis of labels as the goal, and the optimized nursing data mining model is obtained through training;

[0069] The nursing data mining test set is sent to the optimized nursing data mining model for accuracy evaluation. If the evaluation result exceeds the set threshold, the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm is used to optimize the hyperparameters of the initial nursing data mining model until the evaluation result is within the set range to obtain the final nursing data mining model.

[0070] Hyperparameter optimization includes:

[0071] Construct a set of uniformly distributed weight vectors, and use the FBI method to decompose the target optimization problem into a series of single-target sub-problems; randomly generate the initial position vector of the particle, obtain the initial individual best position of the particle, calculate the average best position of the population, and calculate the global optimal position based on the Pareto dominance relationship between the historical individual best positions of different particles;

[0072] Define a set GS, whose size is n(GS), and use the fast non-dominated sorting method to sort the elements in the set GS and all the current individual best position vectors. The higher the non-dominated ranking, the better the quality of the vector. The individual best position vector with the highest ranking is selected and stored in the set GS. All vectors in the set GS are regarded as candidate global optimal solutions and are randomly applied to the next iteration to guide the search of each particle in the population in the target space to obtain the global optimal position.

[0073] Step S5: sending the patient data acquired in real time to the nursing data mining data model for calculation to obtain the nursing data analysis results, and evaluating the patient status and predicting the disease according to the nursing data analysis results;

[0074] Step S6: Obtain the patient's estimated nursing time based on the patient's status assessment and disease prediction results, and dynamically schedule the nurses based on the total estimated nursing time of all patients and nurse data.

[0075] As a preferred technical solution of the present invention, dynamic scheduling includes:

[0076] After obtaining the nursing data analysis results of the patient, match the patient status and disease prediction results in the personalized status assessment strategy library through the patient's personal information, calculate the patient's estimated nursing time based on the matched patient status and disease prediction results, and calculate the total estimated nursing time for all patients;

[0077] Obtain nurse data and dynamically schedule according to each nurse's monthly working hours, weekly working hours, shift tags and levels.

[0078] Embodiment 2 of the present invention is a nursing data automatic collection and analysis system, see Figure 1 and Figure 2 As shown, it includes the following modules:

[0079] The data collection module is used to automatically collect nursing data, including nurse data and patient data. The nurse data is obtained through the nursing staff's work log, including the nursing staff's identity and level, working hours, shift mark and leave data; the patient data is automatically recorded by various sensors equipped in the nursing equipment and filled in by the nursing staff. The automatically recorded patient data includes the patient's body temperature, blood pressure, heart rate, blood oxygen saturation and disease corresponding data. The filled in patient data includes blood sugar measurement value, blood lipid measurement value and disease corresponding data; the acquired patient data is the first patient data;

[0080] The data storage module includes a data preprocessing unit, a change rate calculation unit, a real-time evaluation unit and a storage detection unit. The data preprocessing unit is used to perform data preprocessing on the acquired first patient data, and the preprocessed first patient data is used as the second patient data; the change rate calculation unit is used to calculate the change rate of each data item of the second patient data, and the second patient data is divided into periodic numerical data, non-periodic numerical data and unstructured data according to the data type. The unstructured data is marked with an evaluation category label. For the non-periodic numerical data, the difference percentage of adjacent data is calculated as the change rate. For the periodic numerical data, the root mean square error of adjacent periods is calculated, and the difference percentage between the root mean square error and the mean of the periodic data is calculated. The ratio of the data points is used as the change rate. For unstructured data, the change rate is obtained by changing the category labels of adjacent data. The change rate is recombined with the second patient data to obtain the third patient data. The real-time evaluation unit is used to perform real-time evaluation of the patient data according to the third patient data to obtain the real-time evaluation result of the patient data; the basic pathological information of the patient is obtained. The storage detection unit is used to perform storage detection on the third patient data. The storage detection includes a first storage detection and a second storage detection, which respectively perform data change detection on data points and data segments; the third patient data that passes the storage detection is compressed and stored, and the stored data is used as the fourth patient data. The data in the third patient data that is not completely stored in the fourth patient data is called hidden data;

[0081] A patient data mining module is used to organize the fourth patient data collected historically into a nursing data mining data set, perform data change trend analysis on the nursing data mining data set, add the trend analysis results to the latent data in the nursing data mining data set, use the trend analysis results as data features of the latent data, and use the nursing data mining data set to train an initial nursing data mining model;

[0082] The mining optimization module is used to optimize the category attribute association of the initial nursing data mining model, obtain the optimal pairing parameters of each classification category and the data set of the initial nursing data mining model, and use the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm to optimize the hyperparameters of the initial nursing data mining model. The algorithm adopts a two-stage diversity adjustment mechanism in the search process to avoid the model parameters from falling into the local optimal solution, and obtain the nursing data mining data model;

[0083] The evaluation and prediction module is used to send the real-time acquired patient data into the nursing data mining data model for calculation, obtain the nursing data analysis results, and evaluate the patient status and predict the disease based on the nursing data analysis results;

[0084] The dynamic scheduling module is used to obtain the patient's estimated nursing time based on the patient's status assessment and disease prediction results, and to perform dynamic scheduling based on the total estimated nursing time of all patients and nurse data.

[0085] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for automatic collection and analysis of nursing data, characterized in that: The following steps are involved: Step S1: Automatically collect nursing data, including nurse data and patient data. Nurse data is obtained through the nursing staff's work log, including the nursing staff's identity and level, working hours, shift mark and leave data; patient data is automatically recorded by various sensors equipped in the nursing equipment and filled in by the nursing staff. The automatically recorded patient data includes the patient's body temperature, blood pressure, heart rate, blood oxygen saturation and disease corresponding data. The filled in patient data includes blood sugar measurement value, blood lipid measurement value and disease corresponding data; the acquired patient data is the first patient data; Step S2: performing data preprocessing on the acquired first patient data, and using the preprocessed first patient data as the second patient data; Calculate the change rate for each item of the second patient data, divide the second patient data into periodic numerical data, non-periodic numerical data and unstructured data according to the data type, mark the unstructured data with the evaluation category label, calculate the difference percentage of adjacent data as the change rate for the non-periodic numerical data, calculate the root mean square error of adjacent periods for the periodic numerical data, and use the difference percentage of the root mean square error and the mean of the periodic data as the change rate, obtain the change rate for the unstructured data through the change of the category label of the adjacent data, reorganize the change rate with the second patient data to obtain the third patient data, perform real-time evaluation of the patient data according to the third patient data, and obtain the real-time evaluation result of the patient data; obtain the basic pathological information of the patient, perform storage detection on the third patient data, the storage detection includes a first storage detection and a second storage detection, and respectively perform data change detection on data points and data segments; The third patient data detected by the storage is compressed and stored, and the stored data is used as the fourth patient data, and the data in the third patient data that is not completely stored in the fourth patient data is called hidden data; Step S3: organizing the fourth patient data collected historically into a nursing data mining data set, performing data change trend analysis on the nursing data mining data set, adding the trend analysis results to the latent data in the nursing data mining data set, using the trend analysis results as data features of the latent data, and using the nursing data mining data set to train an initial nursing data mining model; Step S4: Optimize the category attribute association of the initial nursing data mining model, obtain the optimal pairing parameters of each classification category and the data set of the initial nursing data mining model, and use the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm to optimize the hyperparameters of the initial nursing data mining model. The algorithm adopts a two-stage diversity adjustment mechanism in the search process to avoid the model parameters from falling into the local optimal solution, and obtain the nursing data mining data model; Step S5: sending the patient data acquired in real time to the nursing data mining data model for calculation to obtain the nursing data analysis results, and evaluating the patient status and predicting the disease according to the nursing data analysis results; Step S6: Obtain the patient's estimated nursing time based on the patient's status assessment and disease prediction results, and dynamically schedule the nurses based on the total estimated nursing time of all patients and nurse data.

2. The method for automatic collection and analysis of nursing data according to claim 1, characterized in that: Storage testing includes: The average value of the data with a change rate less than 15% in each data of the patient is obtained through the third patient data as the basic status data of the patient, the change rate detection threshold of each data is obtained through the basic status data of the patient, the threshold adjustment weight of the corresponding data is obtained according to the basic pathological information of the patient, and the change rate detection threshold is weightedly calculated to obtain the personalized detection threshold of the patient; a first storage detection, comparing a rate of change of data in the third patient data with a personalized detection threshold, and first marking data exceeding the personalized detection threshold; Second storage detection, using a window of size N to slide and select each item of data in the third patient data, accumulating the change rate of the data in the window, and when the total change rate of the data in the window exceeds N / K*personalized detection threshold, where K is the expected control value of the storage change rate, the first data and the last data in the window are second marked; The data of the first mark and the second mark are compressed and stored.

3. The method for automatic collection and analysis of nursing data according to claim 2, characterized in that: Compressed storage includes: The third patient data is stored in a table format, where each column of the table represents each item of the third patient data. For each item of the third patient data, the data at the first moment is stored, and the data with the first mark and the second mark are stored. The data not stored in the middle is regarded as hidden data. The hidden data is represented by adding a quantity code after the stored data, and its content is consistent with the corresponding stored data. All items of data are merged into a table, and the table is converted into a coded format for storage.

4. The method for automatic collection and analysis of nursing data according to claim 1, characterized in that: The initial nursing data mining model included: The initial nursing data mining model is constructed based on long-short time neural networks. The model includes an input layer for inputting data, L hidden layers for processing data connections, and 1 output layer for outputting classification results. The number of hidden layers in the model is determined by the number of data items in the patient data. The specific number of hidden layers is the number of combinations of three random elements of all data items, the number of classification hidden layers of specific categories, and 1 hidden layer of the final classification result. After the initial model structure is constructed, the model is trained.

5. The method for automatic collection and analysis of nursing data according to claim 4, characterized in that: The training of the initial nursing data mining model includes: The nursing data mining data set is obtained by decoding and restoring the stored fourth patient data, the change trend of the data is obtained according to the adjacent labeled data in the data set, the change trend is reorganized with the latent data, and an analysis label is added to the data corresponding to each patient, and the nursing data mining data set is divided into a training set, a validation set and a test set; The initial nursing data mining model is optimized for category attribute association, and the nursing data mining data set is sent to a pre-trained classification model for calculation to obtain the classification results of the data set, and the weights of each data item corresponding to the classification results in the hidden layer of the model are extracted. The samples corresponding to the same analysis label are statistically analyzed for data weights, and the corresponding data items with the top three weight statistics are obtained as the optimal pairing parameters between the classification category and the data set; if the difference between the top three data in the weight statistics result is too large, the corresponding analysis label is used as a specific category; the analysis parameters are assigned to the hidden layer of the initial nursing data mining model according to the optimal pairing parameters and the specific category; The nursing data mining training set and the validation set are sent to the initial nursing data mining model for training, with the analysis of labels as the goal, and the optimized nursing data mining model is obtained through training; The nursing data mining test set is sent to the optimized nursing data mining model for accuracy evaluation. If the evaluation result exceeds the set threshold, the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm is used to optimize the hyperparameters of the initial nursing data mining model until the evaluation result is within the set range to obtain the final nursing data mining model.

6. The method for automatic collection and analysis of nursing data according to claim 5, characterized in that: Hyperparameter optimization includes: Construct a set of uniformly distributed weight vectors, and use the FBI method to decompose the target optimization problem into a series of single-target sub-problems; randomly generate the initial position vector of the particle, obtain the initial individual best position of the particle, calculate the average best position of the population, and calculate the global optimal position based on the Pareto dominance relationship between the historical individual best positions of different particles; Define a set GS, whose size is n(GS), and use the fast non-dominated sorting method to sort the elements in the set GS and all the current individual best position vectors. The higher the non-dominated ranking, the better the quality of the vector. The individual best position vector with the highest ranking is selected and stored in the set GS. All vectors in the set GS are regarded as candidate global optimal solutions and are randomly applied to the next iteration to guide the search of each particle in the population in the target space to obtain the global optimal position.

7. The method for automatic collection and analysis of nursing data according to claim 1, characterized in that: Dynamic scheduling includes: After obtaining the nursing data analysis results of the patient, match the patient status and disease prediction results in the personalized status assessment strategy library through the patient's personal information, calculate the patient's estimated nursing time based on the matched patient status and disease prediction results, and calculate the total estimated nursing time for all patients; Obtain nurse data and dynamically schedule according to each nurse's monthly working hours, weekly working hours, shift tags and levels.

8. The method for automatic collection and analysis of nursing data according to claim 1, characterized in that: Unstructured data includes: image data, text data, and sound data. The evaluation category labels of unstructured data are divided into major categories and levels, and their change rates are scalar calculated through the differences between major categories and levels.

9. A nursing data automated collection and analysis system, characterized in that: The system applies a method for automatic collection and analysis of nursing data as described in any one of claims 1 to 8, and comprises the following modules: The data collection module is used to automatically collect nursing data, including nurse data and patient data. The nurse data is obtained through the nursing staff's work log, including the nursing staff's identity and level, working hours, shift mark and leave data; the patient data is automatically recorded by various sensors equipped in the nursing equipment and filled in by the nursing staff. The automatically recorded patient data includes the patient's body temperature, blood pressure, heart rate, blood oxygen saturation and disease corresponding data. The filled in patient data includes blood sugar measurement value, blood lipid measurement value and disease corresponding data; the acquired patient data is the first patient data; The data storage module includes a data preprocessing unit, a change rate calculation unit, a real-time evaluation unit and a storage detection unit, wherein the data preprocessing unit is used to perform data preprocessing on the acquired first patient data, and the preprocessed first patient data is used as the second patient data; The change rate calculation unit is used to calculate the change rate of each data item of the second patient data, and divide the second patient data into periodic numerical data, non-periodic numerical data and unstructured data according to the data type. The unstructured data is marked with an evaluation category label. For the non-periodic numerical data, the difference percentage of adjacent data is calculated as the change rate. For the periodic numerical data, the root mean square error of adjacent periods is calculated, and the difference percentage between the root mean square error and the mean of the periodic data is used as the change rate. For the unstructured data, the change rate is obtained by changing the category label of the adjacent data, and the change rate is recombined with the second patient data to obtain the third patient data. The real-time evaluation unit is used to perform real-time evaluation of the patient data according to the third patient data to obtain the real-time evaluation result of the patient data; the basic pathological information of the patient is obtained, and the storage detection unit is used to perform storage detection on the third patient data. The storage detection includes a first storage detection and a second storage detection, which respectively perform data change detection on data points and data segments; The third patient data detected by the storage is compressed and stored, and the stored data is used as the fourth patient data, and the data in the third patient data that is not completely stored in the fourth patient data is called hidden data; A patient data mining module is used to organize the fourth patient data collected historically into a nursing data mining data set, perform data change trend analysis on the nursing data mining data set, add the trend analysis results to the latent data in the nursing data mining data set, use the trend analysis results as data features of the latent data, and use the nursing data mining data set to train an initial nursing data mining model; The mining optimization module is used to optimize the category attribute association of the initial nursing data mining model, obtain the optimal pairing parameters of each classification category and the data set of the initial nursing data mining model, and use the decomposition-based multi-objective quantum behavior particle swarm optimization algorithm to optimize the hyperparameters of the initial nursing data mining model. The algorithm adopts a two-stage diversity adjustment mechanism in the search process to avoid the model parameters from falling into the local optimal solution, and obtain the nursing data mining data model; The evaluation and prediction module is used to send the real-time acquired patient data into the nursing data mining data model for calculation, obtain the nursing data analysis results, and evaluate the patient status and predict the disease based on the nursing data analysis results; The dynamic scheduling module is used to obtain the patient's estimated nursing time based on the patient's status assessment and disease prediction results, and to perform dynamic scheduling based on the total estimated nursing time of all patients and nurse data.

Images