Tumor nursing evaluation method and system based on data mining

Abstract

The invention relates to the technical field of data mining, in particular to a tumor nursing assessment method and system based on data mining, and the method comprises the following steps: based on tumor patient monitoring data, extracting time synchronization data fragments, extracting nursing behavior records, combining the nursing behavior records into combined nursing behavior components, recognizing nursing state transition common features, and carrying out the tumor nursing assessment. And verifying abnormal indexes and behavior response delay, focusing and extracting mutation behaviors, screening state fluctuation associated records, and establishing a dynamic tumor nursing attention sequence. According to the method, the patient monitoring data is finely processed, the time synchronization and standardization of the physiological monitoring data are ensured, the accuracy and integrity of the data are improved, the association of the two data is revealed, the nursing state influence factors are deeply mined, support is provided for a personalized nursing scheme, and the nursing effect is accurately evaluated by identifying the nursing state transfer mode. High-risk behaviors are focused, nursing blind spots are avoided, the nursing efficiency and effect are improved, invalid nursing is reduced, and efficient resource utilization is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data mining, and in particular to a tumor care evaluation method and system based on data mining. BACKGROUND

[0002] The technical field of data mining includes the process of extracting valuable information from a large amount of data, and is widely used in medicine, finance, business and other industries. Data mining analyzes and processes various structured and unstructured data to find potential patterns, relationships and trends between data. The core content includes data preprocessing, feature selection, model establishment, data analysis and pattern recognition, etc. The key of data mining technology lies in how to use statistical methods, machine learning, artificial intelligence, etc. in combination with domain knowledge to discover meaningful information in data. With the rapid development of big data technology, data mining is increasingly widely used in various fields.

[0003] Among them, the tumor care evaluation method refers to an evaluation method based on data mining technology, which aims to assist nursing staff in formulating individualized nursing plans by analyzing relevant nursing data of tumor patients. Through the collection and analysis of multi-dimensional information such as clinical data, quality of life data and treatment feedback data of tumor patients, the nursing needs and treatment effect of patients are evaluated. Through the collection and processing of nursing data, the key factors affecting patient recovery are identified by combining data mining algorithms to help nursing staff carry out effective nursing intervention. This method accurately identifies the nursing focus of patients in a data-driven manner, thereby providing individualized nursing plans. By integrating various types of patient data, a scientific and effective evaluation method for tumor care is provided.

[0004] The prior art relies on traditional data analysis methods when processing patient care data, and there is a lack of deep mining of the relationship between data. In the process of data processing, the multi-dimensional data of patients is not fully synchronized and correlated, making it difficult to fully capture potential key changes in the care process. For example, independent analysis of physiological data and nursing behavior records makes it difficult for nursing staff to quickly identify the relevance between the two, thereby affecting the accuracy and timeliness of nursing decisions. The existing technology is relatively single in detecting changes in nursing status and behavioral responses, lacks comprehensive identification of nursing state transitions, and cannot timely detect mutations or fluctuations in patient status, nor can it dynamically adjust the nursing focus. Therefore, the existing technology cannot flexibly respond to real-time changes in patient status, resulting in information lag and decision-making errors in patient care, ultimately affecting the rehabilitation effect of patients. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and to provide a tumor care evaluation method and system based on data mining.

[0006] In order to achieve the above object, the present application adopts the following technical scheme: A tumor care evaluation method based on data mining, comprising the following steps:

[0007] S1: Based on the daily monitoring data of tumor patients, including the physiological monitoring content of recorded heart rate, body temperature and respiratory rate, extract the time synchronization data segment in the continuous record, classify each data according to the patient identity code, and summarize it into a multi-dimensional list to generate a standard physiological data unit;

[0008] S2: Call the standard physiological data unit, extract the diet, physical recovery, sleep disturbance in the nursing behavior record synchronously, compare the consistency between the two types of data records through time label alignment, and merge into the corresponding patient data structure to obtain the joint nursing behavior group element;

[0009] S3: Based on the joint nursing behavior group element, combined with the sign continuous fluctuation segment, abnormal behavior label, nursing reaction event track, identify the common mode characteristics in the patient nursing state transfer, evaluate and classify the state change node, and obtain the nursing state distribution atlas;

[0010] S4: According to the nursing state distribution atlas, call the label segment with mutation of transfer frequency, cross-verify the group characteristics of abnormal amplitude cluster and behavior reaction delay cluster of monitoring indicators, sort the mutation behaviors through frequency and time span, and form a high-sensitive nursing feature cluster group.

[0011] As a further scheme of the present application, the standard physiological data unit includes heart rate data, body temperature data, respiratory rate data, time synchronization data, patient identity code, the joint nursing behavior group element includes diet record, physical recovery record, sleep disturbance record, consistency comparison result, patient data structure, the nursing state distribution atlas includes nursing state transfer mode, state change node, state distribution characteristics, and the high-sensitive nursing feature cluster group includes mutation label segment, abnormal amplitude cluster, behavior delay cluster, group characteristics, and mutation behavior extraction result.

[0012] As a further scheme of the present application, the acquisition step of the standard physiological data unit is specifically:

[0013] S111: Based on the daily monitoring data of tumor patients, collect and synchronously record the physiological monitoring data of heart rate, body temperature and respiratory rate of patients, arrange them in time sequence and classify them according to patient identity code, analyze the physiological change trend of each patient through data mining, and perform screening to generate physiological data trend analysis result;

[0014] S112: Based on the physiological data trend analysis result, the synchronously recorded physiological data segment is screened, the original data of the patient and the physiological response are analyzed by association, the key factors affecting the physiological change are identified through dynamic change monitoring, and the physiological change influencing factor is generated.

[0015] S113: Based on the physiological change influencing factor, the differential physiological data segment is analyzed by aggregation, cross-validation is performed, and the physiological monitoring data of the patient is optimized, and the standard physiological data unit is generated.

[0016] As a further scheme of the present application, the acquisition step of the joint nursing behavior group element is specifically:

[0017] S211: The standard physiological data unit is called, the diet, physical recovery, and sleep disturbance data in the nursing behavior record are extracted, the data matching at the corresponding time point is compared by aligning the behavior record time, and the behavior synchronization matching marker quantity is generated.

[0018] S212: According to the behavior synchronization matching marker quantity, the time nodes with consistent markers are screened, the corresponding nursing behavior record and physiological data behavior value are extracted, the matching behavior category clustering value is obtained by classifying and aggregating samples by behavior type.

[0019] S213: The matching behavior category clustering value is called, the diet, physical recovery, and sleep disturbance sample record are uniformly identified, and after reordering according to the time label, the formula:

[0020] ;

[0021] is used to calculate the behavior joint mapping value, map and fuse the differential behaviors under the time label, and obtain the joint nursing behavior group element; wherein, represents the behavior joint mapping value, represents the behavior frequency value, represents the number of matched physiological indicators, is the behavior interval density, represents the time point the number of behavior label fluctuations, is the total number of behavior time points.

[0022] As a further scheme of the present application, the acquisition step of the nursing state distribution map is specifically:

[0023] S311: Based on the joint nursing behavior group element and the continuous fluctuation segment of the sign, the behavior fluctuation frequency, sign amplitude, and behavior overlap frequency of each group element are extracted, the sample segment satisfying the sign fluctuation threshold and behavior overlap frequency condition is screened, the coverage density offset value and fluctuation frequency normalized difference value are calculated in combination with the abnormal behavior marker sequence, the samples meeting the determination standard are screened, and the joint fluctuation marker sample value is obtained.

[0024] S312: Call the joint fluctuation mark sample value, fuse the same period nursing reaction event trajectory, calculate the event type distribution density and sample normalization amplitude, match the trajectory density and sample offset difference value, use the formula:

[0025] ;

[0026] Calculate the matching degree interval value, determine the state change node type, and obtain the nursing state distribution map;

[0027] Among them, the matching degree interval value, is the distribution density of the first trajectory, is the normalization amplitude of the first trajectory, is the sample offset difference value of the first trajectory, is the offset amplitude of the first trajectory, is the difference between the sample offset and the reference median of the first trajectory, is the total number of sample categories.

[0028] As a further scheme of the present application, the high-sensitivity nursing feature cluster group acquisition step is specifically:

[0029] S411: According to the nursing state distribution map, call the transition frequency mutation label segment, extract the corresponding abnormal amplitude value cluster and reaction delay value cluster, judge the group distribution in the segment, identify the concentration index, and obtain the mutation group concentration value;

[0030] S412: Call the mutation group concentration value, combine the frequency ranking and span order in the segment, jointly sort the group features, extract the top feature set, and use the formula:

[0031] ;

[0032] Calculate the joint sorting score value, locate the mutation behavior mode through the scoring result, and form a high-sensitivity nursing feature cluster group;

[0033] Among them, the joint sorting score value, is the frequency ranking value of the first sorting group, is the time span order value of the first sorting group, is the average duration of the first sorting group in the label segment, For the first Abnormal fluctuation intensity within the same sorting group, For the total number of sorting groups.

[0034] As a further scheme of the present application, the method further comprises a step S5:

[0035] S5: Based on the high-sensitivity nursing feature cluster group, screening record items associated with state fluctuation, sorting according to the degree of association with risk events, comparing the performance frequency and variation amplitude of additional items within the same period, adjusting the priority weight, and establishing a dynamic tumor nursing attention sequence;

[0036] The dynamic tumor nursing attention sequence includes state fluctuation associated items, risk association sorting, performance frequency, variation amplitude, and priority adjustment weight.

[0037] As a further scheme of the present application, the acquisition step of the dynamic tumor nursing attention sequence is specifically:

[0038] S511: Based on the high-sensitivity nursing feature cluster group, screening record items associated with tumor nursing in the patient nursing data set, calculating the state fluctuation value according to the fluctuation amplitude and performance frequency, evaluating the degree of association of each record item with tumor nursing evaluation, and obtaining tumor nursing evaluation fluctuation record items;

[0039] S512: According to the state fluctuation value of the tumor nursing evaluation fluctuation record item, sorting the record items, comparing the performance frequency and fluctuation amplitude of the differentiated record items within the same period, evaluating the association with risk events, and adjusting the weight value to generate a tumor nursing priority sorting table;

[0040] S513: Based on the tumor nursing priority sorting table, adjusting the priority weight for the dynamic fluctuation trend of each record item, comparing the change frequency of differentiated period record items, and establishing a dynamic tumor nursing attention sequence.

[0041] The tumor nursing evaluation system based on data mining is used to execute the above-mentioned tumor nursing evaluation method based on data mining, and the system comprises:

[0042] The data extraction module extracts the continuous time synchronization data segment of the patient based on the daily monitoring data of the tumor patient, including the physiological monitoring content of the recorded heart rate, body temperature, and respiratory rate, classifies the data according to the patient identity code, and integrates to generate a standard physiological data unit;

[0043] The nursing behavior synchronization module synchronously extracts the diet, physical recovery, and sleep disturbance records associated with patient nursing behavior based on the standard physiological data unit, aligns each type of data through a time tag and performs consistency comparison, and generates a joint nursing behavior group element.

[0044] The state transition mode recognition module recognizes the mode characteristics in the patient care state transition based on the joint care behavior group element, combines the patient sign fluctuation section, the abnormal behavior label sequence and the care response event track, analyzes the change nodes of the care state, and generates a care state distribution atlas;

[0045] The mutation behavior focusing module retrieves the label fragments with abnormal transition frequency based on the care state distribution atlas, analyzes the mutation amplitude of the monitoring indicators and the group characteristics of the behavior reaction delay, sorts them according to the frequency and time span, and forms a high-sensitivity care feature cluster group;

[0046] The dynamic care sequence adjustment module filters the record items associated with the patient state fluctuation based on the high-sensitivity care feature cluster group, combines the performance frequency and the change amplitude within the same cycle, adjusts the priority of the record items, and establishes a dynamic tumor care attention sequence.

[0047] Compared with the prior art, the advantages and positive effects of the present application are:

[0048] In the present application, through the fine processing of the patient daily monitoring data, the accuracy and completeness of the care data can be effectively improved, the time synchronization and coding classification of the physiological monitoring data are ensured, and the standardization of various data items is ensured, thereby providing accurate basic data for further analysis. The synchronization extraction and comparison of care behavior records make the association between care activities and patient physiological data more clear. By combining the two, potential factors affecting the patient care state can be deeply mined, and a comprehensive care behavior data set can be formed to support personalized customization of care programs. By identifying common patterns in care state transition and evaluating state change nodes, the change track of patient care can be fully mastered, and the effect of care intervention can be accurately evaluated. In the process of dynamically adjusting the care attention sequence, the most sensitive care needs of the patient can be efficiently focused, and behaviors and abnormalities with higher risk can be prioritized, thereby avoiding blind spots in the care process, improving care efficiency and effect, and enabling care personnel to more accurately identify patient care needs, improve care quality, optimize patient recovery paths, reduce invalid or repetitive care behaviors, and achieve efficient use of care resources. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 The workflow diagram of the present application;

[0050] Figure 2 The acquisition process diagram of the standard physiological data unit in the present application;

[0051] Figure 3 The acquisition process diagram of the joint care behavior group element in the present application;

[0052] Figure 4 Flow chart for obtaining the nursing state distribution map in the present application;

[0053] Figure 5 Flow chart for obtaining the high sensitivity nursing feature cluster in the present application;

[0054] Figure 6 Flow chart for obtaining the dynamic tumor nursing attention sequence in the present application. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0056] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0057] Example one

[0058] Please refer to Figure 1 The present application provides a technical scheme: a tumor nursing evaluation method based on data mining, comprising the following steps:

[0059] S1: Based on the daily monitoring data of tumor patients, including the physiological monitoring content of recorded heart rate, body temperature, respiratory rate, extract the time synchronization data segment in continuous recording, classify each data according to the patient identity code, and summarize it into a multi-dimensional list, and generate a standard physiological data unit;

[0060] S2: Call the standard physiological data unit, extract the diet, physical recovery, sleep interference in the nursing behavior record synchronously, compare the consistency between the two kinds of data records by time tag alignment, and merge into the corresponding patient data structure to obtain the joint nursing behavior group element;

[0061] S3: Based on the joint nursing behavior group element, combined with the sub-item content of the sign continuous fluctuation segment, abnormal behavior marker sequence, nursing reaction event track, identify the common mode characteristics in the patient nursing state transfer, evaluate and classify the state change node, and obtain the nursing state distribution map;

[0062] S4: According to the nursing state distribution map, the label fragment with transfer frequency mutation is retrieved, the group characteristics of abnormal amplitude cluster and behavior reaction delay cluster of monitoring indicators are cross-verified, the mutation behavior is focused and extracted through the joint sorting of frequency and time span, and a high-sensitivity nursing characteristic cluster group is formed;

[0063] S5: Based on the high-sensitivity nursing characteristic cluster group, the record items highly associated with state fluctuation are screened, the risk events are sorted according to the degree of association, the performance frequency and variation amplitude of additional items in the same cycle are compared, the priority weight is adjusted, and a dynamic tumor nursing attention sequence is established.

[0064] The standard physiological data unit includes heart rate data, body temperature data, respiratory rate data, time synchronization data, patient identity code, the joint nursing behavior group element includes diet record, physical recovery record, sleep disturbance record, consistency comparison result, patient data structure, the nursing state distribution map includes nursing state transition mode, state change node, state distribution characteristic, the high-sensitivity nursing characteristic cluster group includes mutation label fragment, abnormal amplitude cluster, behavior delay cluster, group characteristics, mutation behavior extraction result, and the dynamic tumor nursing attention sequence includes state fluctuation associated item, risk association sorting, performance frequency, variation amplitude, priority adjustment weight.

[0065] Please refer to Figure 2 , the acquisition steps of the standard physiological data unit are as follows:

[0066] S111: Based on the daily monitoring data of tumor patients, the physiological monitoring data of heart rate, body temperature and respiratory rate of patients are collected and recorded synchronously, arranged in time sequence and classified according to patient identity code, the physiological change trend of each patient is analyzed through data mining, and screened to generate physiological data trend analysis result;

[0067] By collecting various physiological monitoring data segments of the patient, a continuous data record is formed, and the data segments are collected by the wearable devices or monitoring tools worn by the patient (such as electrocardiogram monitors, thermometers, respiratory monitors, etc.), and are synchronized with time as the axis to ensure the time consistency of all data segments. For example, the patient's heart rate data is recorded once a minute, the body temperature is recorded once every 15 minutes, and the respiratory rate is recorded once every 30 seconds. The data segments are arranged in chronological order to form a set of continuous monitoring data streams, and all recorded physiological data segments are classified according to the patient's identity code, i.e. the physiological data of each patient is classified separately. This process can be assisted by database technology to identify and store each patient's monitoring data according to a unique patient ID, ensuring that the data of each patient is not confused. The physiological change trend of each patient is analyzed using data mining methods, and by comparing the monitoring data at different time points, the physiological fluctuation trend of the patient can be identified, such as sustained increase in heart rate, gradual increase in body temperature, etc., which helps medical personnel to predict changes in the patient's health status, and the result generates a physiological data trend analysis result, which includes the fluctuation pattern and change trend of the physiological data of each patient, which is helpful for further health assessment and monitoring.

[0068] S112: Based on the physiological data trend analysis result, the synchronously recorded physiological data segments are screened, the patient's original data and physiological response are associated, the key factors affecting physiological changes are identified through dynamic change monitoring, and the physiological change influencing factors are generated;

[0069] First, data screening is performed to remove incomplete or abnormal records. Abnormal records are caused by device failure, sensor error or external environmental interference, such as high temperature or high humidity, which can cause the sensor to produce incorrect readings. For the remaining data segments, correlation analysis is performed in combination with historical records. For example, by comparing the patient's monitoring data over the past few days, the normal fluctuation range of the physiological indicators is analyzed, and the current data is compared with this fluctuation range to determine whether the data change is within the normal fluctuation range or whether an abnormal change has occurred. This process involves data mining techniques such as clustering analysis and trend prediction, which can identify key factors affecting physiological changes, such as symptoms of certain diseases or side effects of medication. Through dynamic change monitoring, factors that have a significant impact on physiological changes can be identified, such as physiological responses after medication adjustment intervention. Comprehensive analysis is performed to obtain the physiological change influencing factors of each patient, which can reveal the underlying reasons for the patient's physiological changes.

[0070] S113: Based on the physiological change influencing factors, the differential physiological data segments are aggregated and analyzed, cross-validated, and the patient's physiological monitoring data is optimized to generate standard physiological data units;

[0071] All physiological monitoring data segments in the multi-dimensional data list are integrated to form a standardized data format, facilitating subsequent analysis and processing. The data processing methods involved include data normalization, missing value filling, etc. to ensure that data of different sources and formats can be consistently integrated together. For example, for heart rate data, there may be different measurement frequencies in different time periods, so frequency unification processing is needed when merging to unify data of different time intervals to one record per minute. Through comparative analysis of different data sets, it is verified whether each item of data conforms to the preset physiological range, and the inconsistency is corrected. The cross-validation process compares the data collected in different time periods to find potential data deviations or abnormalities, further optimizes the data quality, and generates an optimized standard physiological data unit report containing the key physiological data and its change trend of each patient, which is provided to medical staff for reference to make scientific and reasonable nursing decisions. The generation of standard physiological data units is achieved by comprehensively processing the multi-dimensional physiological data of patients to ensure the accuracy and reliability of the data, supporting in-depth analysis and decision support for nursing assessment of tumor patients.

[0072] Please refer to Figure 3 The acquisition steps of the joint nursing behavior group element are as follows:

[0073] S211: Call the standard physiological data unit, extract the diet, physical recovery, and sleep interference data in the nursing behavior record, align the data with the behavior record time, compare the data matching at the corresponding time point, and generate a behavior synchronization matching marker quantity;

[0074] The time tag value of the standard physiological data unit is called and aligned with the behavior occurrence time value in the nursing behavior record. For each aligned time point data, it is checked whether there is consistent behavior record. For example, in actual application scenarios, if a patient eats at a certain time point, its diet record should be synchronized with the corresponding standard physiological data and correspond to the patient's nursing behavior record. If the diet record of a patient is inconsistent with the time tag of its standard physiological data during time alignment, the data is marked as inconsistent, otherwise it is marked as consistent. For time-synchronized data, the recorded behavior content is considered reliable and can be further integrated and analyzed. Based on this alignment and marking process, a set of behavior synchronization matching marker quantities can be obtained for subsequent data screening and analysis. For example, assuming that a patient's diet time record is 12:00, if its standard physiological data also records eating behavior at 12:00, the data is considered consistent, and if its physiological data time is 12:05, it is considered inconsistent, and the behavior synchronization matching marker quantity is obtained.

[0075] S212: According to the behavior synchronization matching label quantity, screen the time nodes with consistent labels, extract the corresponding nursing behavior records and physiological data behavior values, aggregate the samples through behavior type classification, and obtain the matching behavior category clustering value;

[0076] Screen out those time nodes with consistent labels. For each consistent time node, extract the nursing behavior records and standard physiological data unit behavior type values under the time node, and construct a matching sample set according to the information. In actual application, it is assumed that at the time point of 12:00, the patient's nursing records include behaviors of diet, physical recovery and sleep disturbance, and the standard physiological data also records similar behaviors at 12:00. By matching the behavior type values of the two time points, they can be classified into the same behavior category. In the subsequent classification process, the data in the sample set is classified and stored according to the behavior type (for example, diet, physical recovery, sleep disturbance). The data of each behavior category is stored separately, which is convenient for further analysis and processing. The classified sample set will form a matching behavior category clustering value for the next integration analysis.

[0077] S213: Call the matching behavior category clustering value, uniformly identify the diet, physical recovery and sleep disturbance sample records, reorder according to the time label, and use the formula:

[0078] ;

[0079] Calculate the behavior joint mapping value, map and fuse the differential behaviors under the time label, and obtain the joint nursing behavior group element; wherein, represents the behavior joint mapping value, represents the behavior frequency value, represents the physiological index matching quantity, is the behavior interval density, represents the time point the number of label fluctuations, is the total number of behavior time points;

[0080] Reorder according to the time label to ensure the continuity and consistency of time. For each time node, calculate the behavior frequency value , the physiological index matching quantity , the behavior interval density and the label fluctuation number . In order to ensure the dimensional consistency of the parameters, first, the dimensional consistency should be ensured, that is, each parameter should be appropriately standardized or normalized. The frequency value represents the number of times the patient performs the behavior at the time node, and the unit is "times". The physiological index matching quantity The matching degree of physiological data and nursing behavior at this time point, unit: "number", behavior interval density The density of behavior occurrence in a certain time range, unit: "number / minute", label fluctuation number The number of fluctuations of a certain nursing behavior label at this time node, unit: "number", in order to ensure the consistency of each unit, it is necessary to first adjust the dimension of the parameter to a standard unit, normalize all parameters, such as converting to a relative value, which can avoid calculation errors between different dimensions;

[0081] Calculate the above parameters, assuming that at a certain time point (for example, 12:00), the following data is obtained: =2 (the number of eating behaviors at this time point), =5 (the number of matching between physiological data and nursing behavior at this time point), =0.5 (the behavior interval density at this time point), =1 (the number of behavior label fluctuations at this time point), =1 (only one behavior label fluctuation number);

[0082] After substituting into the formula: ;

[0083] The calculation result is the behavior joint mapping value, which represents the joint performance of different behaviors at 12:00. The size of the behavior joint mapping value can reflect the correlation between the behavior characteristics of the patient at this time point and the physiological state. In practical application, through this joint mapping value, various behaviors can be further mapped and fused to obtain joint nursing behavior group elements, and then the nursing scheme can be optimized and the nursing needs of the patient can be more accurately reflected. For example, if the behavior joint mapping value is large, it means that the behavior at this time point is highly consistent with the physiological state, and more frequent intervention and nursing are needed. If the joint mapping value is small, it means that the stability of the behavior at this time point is high, and the demand for nursing intervention is low.

[0084] Please refer to Figure 4 The steps for obtaining the nursing state distribution map are as follows:

[0085] S311: Based on the joint nursing behavior group element and the continuous fluctuation segment of the sign, extract the behavior fluctuation frequency, sign amplitude, and behavior overlap frequency of each group element, filter the sample segments that meet the sign fluctuation threshold and behavior overlap frequency conditions, combine the abnormal behavior label sequence, calculate the coverage density offset value and fluctuation frequency normalized difference value, filter the samples that meet the determination criteria, and obtain the joint fluctuation label sample value;

[0086] Firstly, data screening and classification directly impact the accuracy and effectiveness of the analysis. Data is extracted from the nursing database, with a focus on screening data whose fluctuation amplitude exceeds the set vital sign fluctuation threshold for three consecutive time periods and whose behavioral overlap frequency exceeds the minimum interaction judgment frequency. Taking a real ward monitoring scenario as an example, a nurse observes through monitoring equipment that a patient's heart rate fluctuation is abnormally high over three consecutive time periods, while frequently overlapping with the behavioral patterns of other patients. This indicates a health problem or the need for nursing intervention. The labeled abnormal data is combined with the patient's abnormal behavioral label sequence. By comparing the coverage density, the offset relationship between behavioral patterns can be discovered. For example, if a patient's behavioral labels show a continuous decline in sleep quality, while the data obtained from the monitoring equipment shows abnormal heart rate fluctuations, then this offset in coverage density indicates the need for further medical intervention. By calculating and comparing the difference between the data and the health threshold, the combined fluctuation label sample value is obtained for the individual patient's nursing response indicators.

[0087] S312: Call the joint fluctuation marker sample values, fuse the trajectories of concurrent nursing response events, statistically analyze the event type distribution density and sample normalization amplitude, match the difference between trajectory density and sample offset, and use the following formula:

[0088] ;

[0089] Calculate the matching degree interval value, determine the type of status change node, and obtain the nursing status distribution map;

[0090] in, Indicates the range of matching degree values. For the first Distribution density of the trajectory class For the first Normalized amplitude of the trajectory For the first The difference in sample offsets of the trajectory. For the first The offset magnitude of the trajectory, For the first The difference between the sample offset of the trajectory and the reference median, The total number of sample categories;

[0091] To further characterize the patterns of changes in patient nursing status, it is necessary to integrate sample information with nursing response event trajectories for analysis. First, event trajectory sequences matching the samples within the same time period are selected. The frequency of occurrence of various events within the sample is statistically analyzed, and their density value per unit sample is calculated, denoted as... Using the frequency of events occurring once per minute as the unit of measurement, this ensures the comparability of trajectory frequency data across a unified time scale. For example, if trajectory 1 occurs 15 times within 5 minutes, then... times per minute;

[0092] Based on the distribution value of such trajectory in all matching samples, its fluctuation amplitude is calculated , and it is compressed to the interval [0, 1] by using the normalization method, assuming that the normalized amplitude of trajectory 1 in all samples is ,

[0093] Then, according to the behavior offset of the trajectory sequence in the sample time period, the sample offset difference value is calculated , which is the absolute difference between the actual offset of the behavior pattern and the mean value of the ideal trajectory, such as the offset of sample 1 is 12.3, and the mean value of the trajectory reference is 10.0, then ;

[0094] Then set the trajectory reference offset median value to 9.5, and the deviation of the trajectory from the median value is ;

[0095] In addition, set the ideal value of the trajectory to 10.8, then the trajectory offset reference difference is ;

[0096] To unify the unit, all offset values are measured in "offset unit length", and the unit is the number of behavior marker variations, such as 2.3 represents that the number of offset behavior events is 2.3 times, and if there are trajectories of the same type, the parameters are respectively:

[0097] ;

[0098] ;

[0099] ;

[0100] Substitute the above values into the formula:

[0101] ;

[0102] First calculate the numerator:

[0103] , multiplied by 3.0: 7.254;

[0104] , multiplied by 2.5: 4.578;

[0105] , multiplied by 1.8: 2.367;

[0106] Numerator sum: ;

[0107] Then calculate the denominator: ;

[0108] The final matching degree interval value is: ;

[0109] By merging the trajectory distribution density and the offset difference value, and introducing the relative offset between the trajectory reference value and the sample reference value, the classification matching of abnormal behavior is quantitatively processed, and the accuracy of state node determination is effectively improved. The results show that the current nursing state of the patient has appeared overlapping in multiple abnormal dimensions, and the matching degree greater than 7 indicates that it is in a high-density abnormal event area, and the data needs to be mapped to the corresponding high-density state area in the nursing state distribution map for marking possible state change nodes.

[0110] Referring to Figure 5 , the acquisition steps of the high-sensitivity nursing feature cluster group are:

[0111] S411: According to the nursing state distribution map, the transfer frequency mutation label segment is retrieved, the corresponding abnormal amplitude value cluster and reaction delay value cluster are extracted, the group distribution in the segment is judged, the concentration index is identified, and the mutation group concentration value is obtained;

[0112] According to the nursing state distribution map, the label segment with transfer frequency mutation is retrieved, the state change frequency in the segment is collected, and a state change matrix is constructed in combination with the corresponding time span data. The abnormal amplitude cluster is extracted by inflection point extraction of the continuous change curve of each sign, the number of nodes whose mean difference between the two sides of each mutation node is calculated and exceeds the abnormal amplitude reference value, and at the same time, it is used as the abnormal amplitude count value. The corresponding abnormal amplitude cluster is uniformly normalized in minutes in adjacent segments, and the behavior reaction delay cluster is composed by counting the time difference between the nursing intervention trigger time and the actual nursing response action. The delay value is normalized to a floating point value in the [0, 1] interval in seconds. Then the occurrence frequency of the above abnormal amplitude value cluster and delay value cluster in the label segment is counted, and the threshold is set to 70% quantile of its historical segment. The cluster items higher than the value are selected as candidates, and their distribution characteristics in each label segment are compared in turn. The group characteristics are judged by comparing the group distribution density, arrangement structure and cluster overlap degree. If a cluster item has consistent arrangement order and concentration characteristics in more than two consecutive segments, it is marked as a consistent group. Further, the frequency proportion and time coverage proportion of this type of group in each segment are calculated, and their average value is taken as the concentration index. Finally, the mutation group concentration value is obtained.

[0113] S412: Call the mutation group concentration value, combine the frequency ranking and span sorting in the segment, jointly sort the group characteristics, extract the top feature set, and use the formula:

[0114] ;

[0115] The joint ranking score value is calculated, the mutation behavior pattern is located through the score result, and a high-sensitive nursing feature cluster group is formed.

[0116] Among them, indicates the joint ranking score value, is the frequency ranking value of the first class ranking group, is the time span ranking value of the first class ranking group, is the average duration length of the first class ranking group in the label segment, is the abnormal fluctuation intensity in the first class ranking group, is the total number of ranking groups.

[0117] The frequency ranking value and the time span ranking value thereof in each label segment are obtained, the frequency ranking value is the ranking sequence number of each group in the whole segment according to the frequency from high to low, the time span ranking value is the length ranking position of the covered period of each group, the data acquisition needs to be unified in dimension “minute”, before the joint ranking of the group, the frequency and the span ranking value need to be normalized, the frequency ranking value normalization formula is , wherein indicates the first group ranking, indicates the maximum ranking position, and the span ranking value is the same as , the feature set of the ranking front position interval is obtained and the following operation is performed.

[0118] Among them, the frequency normalization value is set as , the span ranking value is , the duration length value is minute, and the abnormal fluctuation intensity value is , and after substitution, the following is obtained:

[0119] The first item is calculated as: ;

[0120] The second item is: ;

[0121] The third item is: ;

[0122] The sum of the three items is divided by 3: ;

[0123] The formula has the benefit that by simultaneously introducing the four types of participation items of sorting frequency, sorting time span, duration length, and fluctuation intensity, the structure of sum and root is used to measure the comprehensive performance of group sorting characteristics in the time dimension and frequency dimension, avoiding misleading judgments caused by single factor bias. The results show that the group has consistent characteristics in frequency and duration fluctuation, and the comprehensive sorting score is 0.0793, which is in the upper section of the score benchmark interval 0.05-0.10, and can be selected as the focus group accordingly to establish a high-sensitivity nursing feature cluster.

[0124] Please refer to Figure 6 The acquisition steps of the dynamic tumor nursing attention sequence are as follows:

[0125] S511: Based on the high-sensitivity nursing feature cluster, the record items associated with tumor nursing in the patient nursing data set are screened, the state fluctuation value is calculated according to the fluctuation amplitude and performance frequency, the correlation degree of each record item to tumor nursing evaluation is evaluated, and the tumor nursing evaluation fluctuation record item is obtained;

[0126] First, the record items related to tumor nursing are extracted from the patient nursing data set, including the patient's body temperature, heart rate, blood pressure, and other key physiological indicators, and are associated with the patient's tumor type, treatment plan, and other information. In the screening process, all record items directly related to tumor nursing are extracted, which is completed by matching keywords, codes, or through data fields. The "fluctuation amplitude" and "performance frequency" of each record item are calculated. Fluctuation amplitude refers to the change amplitude of the record item within a certain time range, which can be quantified by calculating the standard deviation or the difference between the peak value and the reference value. For example, if a patient's blood pressure fluctuates from 100 / 70 mmHg to 140 / 90 mmHg within a day, the fluctuation amplitude is 40 / 20 mmHg. Performance frequency refers to the frequency of record items appearing within a certain time period. For example, if a patient's heart rate is recorded every hour, and the patient's heart rate is recorded 50 times within 24 hours, the frequency is 50 times / 24 hours. Through these two data, the correlation degree of each record item to tumor nursing evaluation can be evaluated, and the key record items that best reflect the patient's condition changes are selected to obtain the tumor nursing evaluation fluctuation record item. The record item set contains all related record items and their state fluctuation values.

[0127] S512: According to the state fluctuation value of the tumor nursing evaluation fluctuation record item, the record items are sorted, the performance frequency and fluctuation amplitude of the differentiated record items within the same period are compared, the correlation with the risk event is evaluated, and the weight value is adjusted to generate a tumor nursing priority sorting table;

[0128] According to the state fluctuation value, all record items are sorted according to the size of the state fluctuation value, and record items with high fluctuation amplitude or high frequency are given priority. For example, if the heart rate fluctuation amplitude of a patient is large, it indicates that the patient has a high risk, so the record item will be ranked in the front row in the sorting. After sorting, the record items need to be analyzed differently, and the performance frequency and fluctuation amplitude of each record item in the same period are compared for comprehensive evaluation. If the performance frequency or fluctuation amplitude of some record items fluctuates greatly in the period, it means that the record item has high sensitivity in evaluating the patient's condition, and then the relevance to the potential risk event is evaluated. For example, if the body temperature of a patient continues to fluctuate during treatment, and such fluctuation has a strong relevance to the occurrence of tumor treatment side effects (such as infection, complications, etc.), the record item will be considered to be closely related to the risk event. According to the evaluation result, the weight of the record item is adjusted, and a tumor care priority sorting table is generated. Record items with higher weights are given priority in attention and monitoring.

[0129] S513: Based on the tumor care priority sorting table, the priority weight is adjusted according to the dynamic fluctuation trend of each record item, and a dynamic tumor care attention sequence is established by comparing the change frequency of the differentiated period record items;

[0130] According to the tumor care priority sorting table, a preliminary priority weight is assigned to each record item, and the weight is adjusted in real time according to the dynamic fluctuation trend of each record item. For example, when the fluctuation amplitude of a record item (such as blood pressure) significantly increases in a certain period, it indicates that the patient's condition may have changed or risks, so the weight of the record item can be temporarily increased to give more attention. By comparing the change of the performance frequency of the record item in different time periods, it can be evaluated whether the record item still has a high priority. If the fluctuation amplitude and performance frequency of a record item change greatly during subsequent monitoring (such as a previously stable blood pressure becomes fluctuating frequently), its priority needs to be adjusted and its monitoring frequency needs to be increased. Through this dynamic adjustment, a dynamic tumor care attention sequence that changes according to the real-time condition fluctuation of the patient is established, and the sequence is continuously updated as the condition changes, ensuring that the doctor can timely grasp the most urgent care needs of the patient.

[0131] The tumor care evaluation system based on data mining is used to perform the tumor care evaluation method based on data mining described above, and the system comprises:

[0132] The data extraction module extracts continuous time-synchronous data segments of the patient based on the daily monitoring data of the tumor patient, including the physiological monitoring content of the recorded heart rate, body temperature, and respiratory rate, classifies the data according to the patient identity code, and integrates to generate standard physiological data units;

[0133] The nursing behavior synchronization module synchronously extracts diet, physical recovery and sleep interference records associated with patient nursing behaviors based on a standard physiological data unit, aligns each type of data through a time label and performs consistency comparison to generate a joint nursing behavior group element;

[0134] The state transition mode recognition module recognizes mode features in patient nursing state transition, analyzes change nodes of the nursing state and generates a nursing state distribution atlas based on the joint nursing behavior group element in combination with patient sign fluctuation sections, abnormal behavior label sequences and nursing reaction event trajectories;

[0135] The mutation behavior focusing module analyzes mutation amplitude of monitoring indexes and group features of behavior reaction delay, sorts according to frequency and time span, forms a high-sensitivity nursing feature cluster group based on the nursing state distribution atlas and the label fragments with abnormal transition frequency;

[0136] The dynamic nursing sequence adjustment module filters record items associated with patient state fluctuation, adjusts the priority of the record items in combination with performance frequency and variation amplitude within the same cycle, and establishes a dynamic tumor nursing attention sequence.

[0137] The above is only a preferred embodiment of the present application, and does not limit the present application in other forms. Any skilled person in the art can use the disclosed technical content to make changes or modifications to equivalent embodiments applied to other fields. However, any simple modification, equivalent change and modification made to the above embodiments without departing from the technical solution content of the present application, in accordance with the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A tumor care assessment method based on data mining, characterized in that, Includes the following steps: S1: Based on the daily monitoring data of cancer patients, including the recorded physiological monitoring content such as heart rate, body temperature and respiratory rate, extract the time-synchronized data segments in the continuous records, classify each data item according to the patient's identity code, and summarize them into a multidimensional list to generate standard physiological data units. S2: Call the standard physiological data unit, synchronously extract diet, physical recovery, and sleep disturbance from the nursing behavior record, compare the consistency between the two types of data records by time tag alignment, and merge them into the corresponding patient data structure to obtain the joint nursing behavior component; S3: Based on the aforementioned joint nursing behavior components, combined with continuous fluctuations in vital signs, abnormal behavior markers, and nursing reaction event trajectories, identify common pattern features in the transition of patient nursing status, assess and classify status change nodes, and obtain a nursing status distribution map. S4: Based on the nursing status distribution map, retrieve the tag segments with mutations in transfer frequency, cross-validate the group characteristics of abnormal amplitude clusters of monitoring indicators and delayed behavioral response clusters, and extract the mutation behaviors by joint sorting of frequency and time span to form a high-sensitivity nursing feature cluster group; The specific steps for obtaining the high-sensitivity nursing feature clusters are as follows: S411: Based on the nursing status distribution map, retrieve the transfer frequency mutation tag segment, extract the corresponding abnormal amplitude value cluster and response delay value cluster, determine the group distribution in the segment, identify the concentration index, and obtain the mutation group concentration value. S412: Using the concentration value of the mutation group, combined with the frequency ranking and span ranking in the fragments, the group features are jointly ranked, and the top-ranked feature set is extracted using the following formula: ； Calculate the joint ranking score, locate the mutation behavior pattern through the score results, and form a cluster of highly sensitive nursing characteristics; in, This represents the joint ranking score. For the first The frequency ranking value of the sorted group. For the first Sort values ​​by time span for sorted groups. For the first The average duration of the sorted group within the label fragment. For the first Intensity of abnormal fluctuations within the sorted group This represents the total number of sorted groups; S5: Based on the high-sensitivity nursing feature clusters, filter the record items that are highly correlated with the state fluctuations, sort them according to the degree of correlation of risk events, compare the performance frequency and change of additional items in the same period, adjust the priority weights, and establish a dynamic tumor nursing attention sequence. The dynamic tumor care attention sequence includes status fluctuation correlation items, risk correlation ranking, performance frequency, change magnitude, and priority adjustment weight.

2. The tumor nursing assessment method based on data mining according to claim 1, characterized in that, The standard physiological data unit includes heart rate data, body temperature data, respiratory rate data, time synchronization data, and patient identification code. The joint nursing behavior unit includes dietary records, physical recovery records, sleep disturbance records, consistency comparison results, and patient data structure. The nursing status distribution map includes nursing status transition patterns, status change nodes, and status distribution characteristics. The high-sensitivity nursing feature clusters include mutation label fragments, abnormal amplitude clusters, behavioral delay clusters, group features, and mutation behavior extraction results.

3. The tumor nursing assessment method based on data mining according to claim 1, characterized in that, The specific steps for acquiring the standard physiological data unit are as follows: S111: Based on daily monitoring data of cancer patients, collect and synchronously record physiological monitoring data of patients' heart rate, body temperature and respiratory rate, arrange them in chronological order and classify them according to patient identity code, analyze the physiological change trend of each patient through data mining, and filter them to generate physiological data trend analysis results. S112: Based on the physiological data trend analysis results, screen synchronously recorded physiological data segments, analyze the correlation between the patient's original data and physiological responses, identify key factors affecting physiological changes through dynamic change monitoring, and generate factors influencing physiological changes; S113: Based on the factors influencing the physiological changes, differentiated physiological data fragments are aggregated and analyzed, cross-validated, and the patient's physiological monitoring data is optimized to generate standard physiological data units.

4. The tumor nursing assessment method based on data mining according to claim 3, characterized in that, The specific steps for obtaining the combined nursing behavior components are as follows: S211: Call the standard physiological data unit to extract the diet, physical recovery and sleep disturbance data in the nursing behavior record, align it with the behavior record time, compare the data matching at the corresponding time point, and generate behavior synchronization matching marker quantity; S212: Based on the behavior synchronization matching marker quantity, filter time nodes with consistent markers, extract corresponding nursing behavior records and physiological data behavior values, and aggregate samples through behavior type classification to obtain matching behavior category cluster values; S213: Call the clustering values ​​of the matched behavior categories to uniformly identify the records of diet, physical recovery, and sleep disturbance samples. After re-sorting by time label, use the following formula: ； Calculate the joint behavioral mapping value, and map and fuse the differentiated behaviors under time tags to obtain the joint nursing behavior component; among which... Indicates the behavior-joint mapping value. Represents the frequency of behavior. The number of matching physiological indicators. For behavioral interval density, Representing a point in time The next line shows the number of label fluctuations. This represents the total number of time points in the action.

5. The tumor nursing assessment method based on data mining according to claim 4, characterized in that, The specific steps for obtaining the nursing status distribution map are as follows: S311: Based on the combined nursing behavior components and continuous fluctuation segments of vital signs, extract the frequency of behavioral fluctuations, amplitude of vital signs, and frequency of behavioral overlap for each component, screen sample segments that meet the threshold of vital sign fluctuations and the condition of frequency of behavioral overlap, combine the abnormal behavior marker sequence, calculate the coverage density offset value and the normalization difference of fluctuation frequency, screen samples that meet the judgment criteria, and obtain the combined fluctuation marker sample value. S312: Call the aforementioned joint fluctuation marker sample values, fuse the synchronous nursing response event trajectories, statistically analyze the event type distribution density and sample normalization amplitude, match the difference between trajectory density and sample offset, and use the formula: ； Calculate the matching degree interval value, determine the type of status change node, and obtain the nursing status distribution map; in, Indicates the range of matching degree values. For the first Distribution density of the trajectory class For the first The normalized magnitude of the trajectory. For the first The difference in sample offsets of the trajectory. For the first The offset magnitude of the trajectory, For the first The difference between the sample offset of the trajectory and the reference median, This represents the total number of sample categories.

6. The tumor nursing assessment method based on data mining according to claim 1, characterized in that, The specific steps for obtaining the dynamic tumor care attention sequence are as follows: S511: Based on the high-sensitivity nursing feature cluster, screen the records in the patient nursing data that are associated with tumor nursing, calculate the state fluctuation value according to the fluctuation amplitude and the frequency of performance, evaluate the correlation between each record item and the tumor nursing assessment, and obtain the tumor nursing assessment fluctuation record item. S512: Based on the status fluctuation values ​​of the tumor care assessment fluctuation record items, sort the record items, compare the frequency and fluctuation amplitude of the differential record items in the same period, assess the correlation with risk events, adjust the weight values, and generate a tumor care priority ranking table. S513: Based on the aforementioned tumor care priority ranking table, adjust the priority weight for each record item according to its dynamic fluctuation trend, and establish a dynamic tumor care attention sequence by comparing the change frequency of record items in different periods.

7. A tumor care assessment system based on data mining, characterized in that, The system is used to implement the data mining-based tumor care assessment method according to any one of claims 1-6, and the system comprises: The data extraction module is based on the daily monitoring data of cancer patients, including the recorded physiological monitoring content such as heart rate, body temperature and respiratory rate. It extracts the continuous time-synchronized data segments of patients, classifies the data according to the patient identification code, and integrates them to generate standard physiological data units. The nursing behavior synchronization module, based on the standard physiological data unit, synchronously extracts records of diet, physical recovery, and sleep disturbances associated with patient nursing behaviors, aligns each type of data through time tags and performs consistency comparison to generate joint nursing behavior components; The state transition pattern recognition module, based on the joint nursing behavior components, combines the patient's vital sign fluctuation segments, abnormal behavior marker sequences, and nursing response event trajectories to identify pattern features in the patient's nursing state transitions, analyze the change nodes of nursing state, and generate a nursing state distribution map. Based on the nursing status distribution map, the mutation behavior focusing module retrieves the tag segments with abnormal transfer frequency, analyzes the mutation magnitude of the monitoring indicators and the group characteristics of behavioral response delay, and sorts them according to frequency and time span to form high-sensitivity nursing feature clusters. The dynamic nursing sequence adjustment module, based on the aforementioned high-sensitivity nursing feature clusters, filters record items associated with fluctuations in patient status, and adjusts the priority of record items by combining the frequency and magnitude of changes within the same period, thereby establishing a dynamic tumor nursing attention sequence.

Images