Method for automatically generating and distributing pension nursing tasks

Abstract

The invention relates to the technical field of old-age nursing task automatic generation and distribution, and discloses an old-age nursing task automatic generation and distribution method. The overall process comprises the steps of collecting basic data of the old people and nursing personnel, defining original variables, performing standardization processing on the data, and calculating feature distances between the old people; then adaptively determining distance scale parameters, constructing a similarity matrix and extracting connected components, and realizing demand clustering of the elderly; and on the basis, a greedy strategy is adopted to select the least nursing personnel to cover the service demand, tasks are accurately distributed according to the capability proportion, and a result is output. According to the method, a whole-process closed loop from data acquisition to task allocation is realized, and the method has an end-to-end automation capability. Through standardized modeling, similarity calculation, graph theory clustering and a greedy optimization algorithm, the scheme realizes intelligent nursing management in a resource limited environment.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic generation and allocation of elderly care tasks, and specifically to a method for automatic generation and allocation of elderly care tasks. Background Art

[0002] With the acceleration of the aging population process, elderly care institutions and home care service systems are facing multiple challenges such as the matching of caregivers' supply and demand, personalized services, and optimization of human resources. Traditional allocation of elderly care tasks mainly relies on the experience of managers and manual scheduling, usually grouping and allocating according to the time of elderly people's admission, bed location, or simple demand types (such as basic care, rehabilitation escort). This method has problems such as uneven workload, large fluctuations in service quality, and neglect of individual differences among the elderly.

[0003] In the prior art, some institutions have introduced scheduling software to automatically generate caregiver schedules by setting fixed shifts and upper and lower limits of working hours. However, such software only performs linear scheduling for the daily total working hours constraint and does not refine to the level of individual elderly people, nor does it consider the matching degree of the elderly people's behavior characteristics, health status, and caregivers' professional skills. In addition, some studies have tried to use fuzzy rules or heuristic algorithms to match the elderly people and caregivers, but mostly rely on manually set weight parameters. These weights lack a unified standard and cannot adapt to data changes, making it difficult to ensure the objectivity and fairness of the allocation scheme. With the popularization of the Internet of Things and wearable devices, daily activity data of the elderly such as the number of steps, sleep duration, and toilet visits can be obtained in real time through smart bracelets or mattress sensors. However, currently only a few systems use these multi-dimensional dynamic data for care task allocation, mostly simply counting health indicators or giving risk assessments in the offline analysis stage, failing to deeply explore the similarities and differences among the elderly, nor forming a continuous closed-loop automatic scheduling process. On the other hand, graph theory and clustering techniques have been maturely applied in fields such as social network analysis and logistics routing optimization. For example, the graph partitioning method based on community discovery can automatically divide a large number of nodes into several relatively independent subgroups; the threshold selection strategy based on the median or mean can achieve static segmentation of the graph structure. However, these technologies mostly remain in the academic exploration stage in elderly care allocation, have not been closely combined with care workload, care duration, and real-time needs, nor formed a full-process algorithm from data collection, feature calculation, to connected component extraction and then to resource allocation.

[0004] To this end, the present case aims to propose an automatic generation and allocation method for elderly care tasks. By modeling and clustering analysis of the multi-dimensional behavioral data of the elderly, combined with the ability parameters of the caregivers, a data-driven intelligent task allocation mechanism is constructed. The scheme fully integrates data collection, standardized processing, similarity construction, graph theory modeling and greedy optimization strategies, and can scientifically group the elderly according to their similar care needs, and on this basis, optimize the matching of personnel and tasks. Summary of the Invention

[0005] The present invention provides an automatic generation and allocation method for elderly care tasks, which promotes the solution of the problems mentioned in the above background art.

[0006] The present invention provides the following technical solutions: an automatic generation and allocation method for elderly care tasks, including:

[0007] Collect the basic data of the elderly and the caregivers, and define the original variables, including the daily behavior characteristics of the elderly and the service ability parameters of the caregivers;

[0008] Standardize the original feature data of the elderly to form a standardized feature vector;

[0009] Calculate the feature distance between any two elderly based on the standardized features;

[0010] Calculate the distance scale parameter according to the feature distance and make an adaptive adjustment;

[0011] Construct a similarity matrix based on the distance scale parameter and calculate the similarity threshold;

[0012] Use the similarity matrix to construct an undirected graph and extract the connected components;

[0013] Select a subset of caregivers that meet the needs of the elderly based on the greedy strategy;

[0014] Allocate care tasks according to the needs of the elderly and the ratio of caregiver abilities, and output the task allocation result.

[0015] Optionally, the collection of the basic data of the elderly and the caregivers and the definition of the original variables specifically include:

[0016] Let the total number of the elderly be , the index set of the elderly be , the index corresponds to the th elderly person, denoted as ;

[0017] Let the total number of the caregivers be , the index set of the caregivers be , the index corresponds to the One caregiver, denoted as ;

[0018] For each elderly person Collect five - dimensional original features , including:

[0019] , average daily walking steps in the past 7 days; , average daily sleep duration in the past 7 days; , average daily toilet use times in the past 7 days; , average daily meal times in the past 7 days; , average daily medication times in the past 7 days; where is the original feature index;

[0020] Let the daily care requirement duration of the elderly person be , and the available service duration of the caregiver on the same day be .

[0021] Optionally, standardize the original feature data of the elderly person to form a standardized feature vector, specifically including:

[0022] Calculate the arithmetic mean of the -th dimensional original feature;

[0023] Calculate the standard deviation of the -th dimensional original feature;

[0024] Construct the standardized feature vector of the elderly person , ; where is the standardized value of the -th dimension.

[0025] Optionally, calculate the feature distance between any two elderly people based on the standardized features, specifically including:

[0026] For any two elderly person indices , calculate their Euclidean distance :

[0027] .

[0028] Optionally, calculate the distance scale parameter based on the feature distance and perform adaptive adjustment, specifically including:

[0029] Set the overall distance scale parameter to .

[0030] Optionally, constructing a similarity matrix based on the distance scale parameter and calculating a similarity threshold specifically includes:

[0031] Set the demand similarity between the elderly and to be , ;

[0032] Construct the similarity set as ;

[0033] Calculate the median threshold ; where is the median function, and the median value is taken after sorting .

[0034] Optionally, constructing an undirected graph using the similarity matrix and extracting connected components specifically includes:

[0035] Construct an adjacency indicator function:

[0036] ; where is an adjacency indicator variable, which takes 1 if and only if the similarity between two elderly people is not less than the threshold , indicating that an edge should be connected between vertices and in the undirected graph; otherwise it is 0 and no edge is connected;

[0037] Initialize the set of unvisited vertex indices , and the connected component counter ;

[0038] When , repeat the following steps:

[0039] S1100. Let ;

[0040] S1200. Take the smallest index from , set the current component vertex set , and delete from ;

[0041] S1300. Set the pointer set ;

[0042] S1400. When :

[0043] S1410. Arbitrarily select , and delete it from ;

[0044] S1420. For all check, if , then execute:

[0045] S1421. Add to ;

[0046] S1422. Add to ;

[0047] S1423. Delete from ;

[0048] S1500. Until is reached, is the vertex set of the th connected component. Denote its edge set as:

[0049] ;

[0050] S1600. Repeat steps S1100 to S1500 until ;

[0051] Finally, obtain the connected component set .

[0052] Optionally, the selection of the subset of caregivers meeting the needs of the elderly based on the greedy strategy specifically includes:

[0053] For each component execute the following steps:

[0054] S201. Calculate the total care needs of all the elderly in the component ;

[0055] S202. Sort the indices in in descending order according to . Denote the sorted sequence as , satisfying ; where is the index of the th caregiver with the largest available duration;

[0056] S203. Let , and calculate the cumulative available duration ; where is a counter used to record the number of members currently selected into the subset of caregivers;

[0057] If , let and repeat; otherwise stop;

[0058] S204. Select the smallest set of caregivers that meets the component demand , wherein, is the total available duration of the set .

[0059] Optionally, according to the ratio of the needs of the elderly to the capabilities of the caregivers, allocate care tasks and output the task allocation result, specifically including:

[0060] For each and each caregiver calculate: wherein, is the service duration of the elderly assigned to the caregiver in the component ;

[0061] Verify:

[0062] S301. Satisfaction of the needs of the elderly: ;

[0063] S302. Non-overloading of caregivers: ;

[0064] For each component output:

[0065] List of grouped elderly , subset of caregivers and allocation matrix .

[0066] The present invention has the following beneficial effects:

[0067] 1. The solution proposed in the scheme comprehensively reflects the daily living conditions of the elderly from five dimensions (walking, sleeping, toileting, dining, taking medicine), which is far better than the existing static demand collection methods based only on questionnaires or subjective ratings. This method of feature definition based on time-series behavior data has the characteristics of high resolution and quantifiability, and can accurately reflect the actual care intensity requirements of the elderly. Most traditional systems rely on care level classification and lack fine-grained data support, making it difficult to achieve differentiated services. This solution constructs a multi-dimensional and structured input space, providing a solid data foundation for subsequent algorithms.

[0068] 2. By calculating the mean and standard deviation, the features of all dimensions are normalized to a zero-mean and unit-variance space, solving the problem of incomparability between features of different dimensions. This step significantly improves the accuracy and fairness of subsequent distance calculations and is a prerequisite for standardization in data mining and machine learning algorithm applications. The traditional way of allocating pension tasks mostly uses a scoring system, which has problems such as strong subjectivity and inconsistent dimensions. This solution constructs an objectively comparable feature space through mathematical normalization, making the model input more logical and consistent.

[0069] 3. The Euclidean distance is used to measure the comprehensive differences between the elderly, and a distance scale adaptive mechanism based on the global root mean square is further proposed. This mechanism does not rely on the setting of empirical thresholds and can automatically adjust the similarity conversion scale according to the actual data distribution, effectively improving the generality and robustness of the model. In existing research, similarity calculations often rely on fixed distance scales and are easily interfered by outliers. This solution uses a statistical adaptive method to automatically adapt to the differences in data sets under different scenarios, significantly enhancing the environmental adaptability of the system.

[0070] 4. The Euclidean distance is converted into a similarity score in the range of 0 to 1, and the median is used as the threshold to avoid bias introduced by manually setting parameters. This dynamic threshold method based on the statistical median can achieve efficient clustering with tightness within the group and separation between groups. Traditional methods often rely on empirical thresholds or preset clustering K values and do not have self-adaptability. The data-driven partitioning strategy of this method significantly improves the scientificity and rationality of the elderly grouping.

[0071] 5. The elderly with a similarity higher than the threshold are connected by edges in an undirected graph, and a connected component extraction algorithm is used to automatically identify groups of the elderly with similar care needs. This graph theory modeling method is both efficient and intuitive and can clearly define the task allocation unit. Most existing allocation systems only conduct single-person evaluations and ignore the group attributes among the elderly. This solution starts from the "group-caregiver" dimension and significantly improves the aggregation degree and accuracy of task grouping.

[0072] 6. On the basis of grouping, a greedy algorithm is introduced to gradually cover the needs starting from caregivers with high service capabilities until the minimum set is reached. This heuristic selection method takes into account both efficiency and resource optimization and is especially suitable for environments with tight resources. Common scheduling systems mostly rely on rotation or equal division strategies and ignore the matching of task loads and human resources. This solution realizes on-demand dynamic scheduling, greatly improving the utilization rate and flexibility of nursing resources.

[0073] 7. Introduce a "proportional" allocation mechanism to ensure that the service volume of caregivers is directly proportional to their available working hours, and add a validation of the legality of the allocation results (demand satisfaction and non-overloading). The task output is structured as "cluster list + caregiver subset + allocation matrix", which can be directly connected to the scheduling system. The allocation mechanism of the traditional system is extensive and does not support many-to-many mapping and refined task output. This solution has a complete mathematical basis and output structure, and is the core support for the intelligent upgrade of the nursing scheduling system. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 It is a schematic flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0076] Embodiment, referring to Figure 1 , a method for automatically generating and allocating elderly care tasks, including:

[0077] Collect the basic data of the elderly and caregivers, and define the original variables, including the daily behavior characteristics of the elderly and the service ability parameters of the caregivers;

[0078] Standardize the original feature data of the elderly to form a standardized feature vector;

[0079] Calculate the feature distance between any two elderly people based on the standardized features;

[0080] Calculate the distance scale parameter according to the feature distance and perform adaptive adjustment;

[0081] Construct a similarity matrix based on the distance scale parameter and calculate the similarity threshold;

[0082] Construct an undirected graph using the similarity matrix and extract the connected components;

[0083] Select a subset of caregivers that meet the needs of the elderly based on the greedy strategy;

[0084] Allocate the care tasks according to the needs of the elderly and the proportion of caregiver capabilities, and output the task allocation results.

[0085] First, collect the behavior and ability data of the elderly and caregivers to provide the basic input for subsequent model calculations. Then, through data processing means such as feature standardization, similarity measurement, and graph construction, complete the clustering of the elderly's demand characteristics. Next, select nursing resources through the greedy algorithm and accurately allocate nursing tasks according to the ability ratio. This method realizes a closed-loop from "data collection" to "task output" and has end-to-end automation capabilities. This solution solves the problems of rough, unreasonable, and slow response in task allocation in traditional elderly care through standardized modeling, graph theory partitioning, and heuristic optimization. The system can intelligently identify similar nursing demand groups and optimally allocate caregivers under resource constraints, realizing the dynamic optimization and refined management of nursing tasks, improving service efficiency and fairness, and having high practical value.

[0086] Collect the basic data of the elderly and caregivers and define the original variables, specifically including:

[0087] Let the total number of the elderly be , and the set of elderly indexes be , where the index corresponds to the th elderly person, denoted as ; identify the number of the elderly in the system and its number range to ensure the orderliness of subsequent processing;

[0088] Let the total number of caregivers be , and the set of caregiver indexes be , where the index corresponds to the th caregiver, denoted as ; identify the number of caregivers and its number range to support the ability allocation process;

[0089] For each elderly person collect five-dimensional original features , including:

[0090] , the average number of daily walking steps in the past 7 days; , the average daily sleep duration in the past 7 days; , the average number of daily toilet visits in the past 7 days; , the average number of daily meals in the past 7 days; , the average number of daily medication times in the past 7 days; where is the original feature index; provide quantitative demand and behavior indicators and input the original data for subsequent similarity calculation;

[0091] Let the elderly person's nursing demand duration on the current day be , and the caregiver's available service duration on the current day be Quantify the demand of the elderly and the available resources of caregivers, and construct the allocation boundary.

[0092] Specific variable structures of "collecting original data" are defined in detail, including the number of the elderly, the number of caregivers, five-dimensional behavioral characteristics, care needs, and available service time. By establishing a unified data index set and variable system, the standardization of the model input data and the orderliness of calculations are ensured. By systematically collecting five core behavioral characteristics, this method establishes a solid data foundation for subsequent similarity calculation and graph clustering at the quantification level. Compared with the traditional subjective evaluation-based care demand division method, this technology has significant advantages such as objectivity, real-time, and scalability, laying a foundation for data-driven intelligent scheduling and improving the modeling accuracy and system stability.

[0093] The original feature data of the elderly is standardized to form a standardized feature vector, which specifically includes:

[0094] Eliminate the influence of different dimensions, and map the five-dimensional features to a unified dimensionless space;

[0095] Calculate the arithmetic mean of the -dimensional original feature ; Obtain the data center position of this dimension for debiasing;

[0096] Calculate the standard deviation of the -dimensional original feature ; Measure the data dispersion degree of this dimension for scaling;

[0097] Construct the standardized feature vector of the elderly , ; Among them, is the standardized value of the -dimension; Normalize the original features to zero mean and unit variance to ensure the equal importance of each dimension.

[0098] Standardize the five-dimensional original features, that is, uniformly convert data with different dimensions into a standard normal distribution space with zero mean and unit variance. This process effectively eliminates the influence of dimensions between feature dimensions, making the subsequent Euclidean distance calculation more reasonable and fair. Through standardization, the problem of "a certain dimension dominating the distance calculation due to large values" is avoided. This operation ensures the equal weight of each dimension in the calculation, improves the model's ability to identify the true differences in the care needs of the elderly, enhances the scientificity and discriminability of the feature space, and is a key technical step to ensure the accuracy of clustering and allocation.

[0099] The feature distance between any two elderly is calculated based on the standardized features, which specifically includes:

[0100] The original metric for measuring the similarity of features among the elderly using Euclidean distance;

[0101] For any two elderly indexes , calculate the Euclidean distance between them :

[0102] ;

[0103] Quantify the overall difference in the standardized features of two elderly people as the basis for subsequent similarity conversion.

[0104] Using Euclidean distance to calculate pairwise for the standardized feature vectors to measure the similarity of care needs among the elderly is the basis for similarity calculation and subsequent graph construction. As the basic form of measuring the difference between continuous variables, the introduction of Euclidean distance provides an intuitive and controllable quantitative index for the system to construct the "Elderly Demand Difference Model", enabling nursing homes to more accurately judge the service similarity of the elderly, solving the problem that the traditional "one-size-fits-all" grouping method cannot adapt to individual differences, and improving the degree of service personalization.

[0105] The distance scale parameter is calculated according to the feature distance and adaptively adjusted, specifically including:

[0106] Define the scaling factor through the full-distance data to ensure a reasonable numerical range of the similarity function;

[0107] Set the overall distance scale parameter to ;

[0108] Take the root mean square of all distances to automatically adapt to the data dispersion without manual setting.

[0109] Calculate the root mean square of all distance values as the distance scaling factor to achieve the scale adaptive adjustment of the feature space. This mechanism can prevent the distortion of distance calculation caused by uneven distribution of input data. By globally adapting the similarity scale with the root mean square, this method avoids the adaptability problems brought by static parameters, improves the generality and stability of the model. Especially in the face of scenarios with different scales of nursing homes and different data sampling frequencies, it can still ensure the consistency and effectiveness of the clustering logic, greatly enhancing the robustness and cross-environment deployment ability of the model.

[0110] The similarity matrix is constructed based on the distance scale parameter and the similarity threshold is calculated, specifically including:

[0111] Map the distance to the similarity within the interval and adaptively select the division threshold;

[0112] Set the demand similarity between the elderly and to be , Convert the distance value to similarity, where the smaller the distance, the closer the similarity is to 1;

[0113] Construct a similarity set as Collect all non-self-loop similarities for statistical analysis;

[0114] Calculate the median threshold ; where is the median function, which takes the median value after sorting ; Select the median of the similarities as the division threshold to ensure balance within and between groups.

[0115] Map the Euclidean distance to the similarity between and determine the division threshold based on the median of all similarities, so as to ensure high similarity among the elderly within the group and low similarity between groups. By using the median instead of an empirical value as the clustering threshold, the objectivity and robustness of the division are improved, and the problem of difficult to set the clustering boundary is solved, especially suitable for automatic classification under large-scale data. Compared with static clustering algorithms such as K-means, it does not require presetting the number of clusters and can more truly reflect the natural distribution relationship of the nursing needs among the elderly.

[0116] The construction of an undirected graph using the similarity matrix and the extraction of connected components specifically include:

[0117] Construct a similarity threshold graph and accurately extract connected components;

[0118] Construct an adjacency indicator function:

[0119] ; where is an adjacency indicator variable, which takes 1 if and only if the similarity between two elderly people is not less than the threshold , indicating that an edge should be connected between vertices and in the undirected graph; otherwise it is 0 and no edge is connected; Determine whether an edge connection is established between two elderly people in the graph;

[0120] Initialize the set of unvisited vertex indices , and the connected component counter ; Prepare the set of unvisited nodes and the connected component counter;

[0121] When , repeat the following steps:

[0122] S1100. Let ; Start the extraction of the th connected component;

[0123] S1200. Take the smallest index from ​ , set the current component vertex set , and delete it from ; Initialize the component vertex set;

[0124] S1300. Set the pointer set ; Used to iteratively expand vertices;

[0125] S1400. When :

[0126] S1410. Arbitrarily select , and delete it from ;

[0127] S1420. Check all . If , then execute:

[0128] S1421. Add to ;

[0129] S1422. Add to ;

[0130] S1423. Delete from ;

[0131] Expand the component vertices in a breadth - first manner until the component is fully formed;

[0132] S1500. Until , is the vertex set of the th connected component. Denote its edge set as:

[0133] ; Record the legal edges between all vertices within this component;

[0134] S1600. Repeat steps S1100 to S1500 until ; Extract all connected components;

[0135] Finally, obtain the connected component set .

[0136] Construct an undirected graph based on the similarity matrix, and then obtain the elderly groups with similar needs through the connected component extraction algorithm. Transforming data into a structure by means of graph theory enhances the information organization ability. The graph model can better retain complex relationships and local structure information compared with traditional clustering methods. The extraction of connected components realizes the structured grouping of nursing tasks, effectively supports the local optimization and parallel processing of subsequent task allocation, improves the response efficiency of nursing resource scheduling and the algorithm scalability, and is a highly cutting-edge modeling method in group management.

[0137] The above-mentioned selection of a subset of caregivers to meet the needs of the elderly based on the greedy strategy specifically includes:

[0138] Within each group of the elderly, meet the total demand with the fewest caregivers;

[0139] For each component Execute the following steps:

[0140] S201. Calculate the total nursing demand of all the elderly in the component ; count the total nursing demand of all the elderly under this component;

[0141] S202. Sort the indices in in descending order according to , and record the sorted sequence as , satisfying ; where is the index of the caregiver with the -th largest available duration; prepare the basis for greedy selection of people;

[0142] S203. Let , and calculate the cumulative available duration ; where is a counter used to record the number of members currently selected into the subset of caregivers;

[0143] If , let Repeat; otherwise stop;

[0144] Add the caregiver with the highest available duration one by one until the demand is covered;

[0145] S204. Select the smallest set of caregivers that meets the demand of the component , ; where is the total available duration of the set ; output the smallest set of caregivers required for this component and its total resource volume.

[0146] Within each subgroup, a greedy strategy is adopted to select the minimum number of caregivers to cover all the needs of the elderly. Caregivers with longer service hours are given priority to improve the efficiency of resource utilization. The greedy strategy minimizes resource input while meeting the coverage requirement, not only reducing the number of caregivers needed but also ensuring the task fulfillment rate. This mechanism of "meeting the maximum demand with the minimum resources" is particularly suitable for current nursing homes with high labor costs and tight resource allocation, greatly optimizing the resource allocation structure and operating costs.

[0147] Allocate nursing tasks according to the ratio of the needs of the elderly to the capabilities of the caregivers, and output the task allocation results, which specifically include:

[0148] Allocate specific durations within each subgroup according to the ratio of caregiver capabilities, and verify that there is no overloading and no surplus;

[0149] For each and each caregiver calculate: ; where is the service duration of the elderly in component assigned to caregiver ; Allocate the needs of the elderly according to the ratio of the available duration of the caregivers to ensure fairness in the ratio;

[0150] Verify:

[0151] S301. Satisfaction of the needs of the elderly: ;

[0152] S302. Non-overloading of caregivers: ;

[0153] Ensure legal allocation, all the elderly receive services, and the load of caregivers does not exceed the limit;

[0154] For each component output:

[0155] List of subgroups of the elderly , subset of caregivers and allocation matrix ; for direct call and execution by the scheduling system or management personnel.

[0156] Finally, allocate the nursing tasks according to the ratio of the available duration of the caregivers, and set up a dual-constraint verification mechanism (satisfaction of the needs of the elderly and non-overloading of caregivers) to ensure the fairness and legality of the task allocation. Through proportional allocation and constraint verification, the feasibility of task allocation and the safety of resource utilization are guaranteed, effectively preventing the problems of overloading of caregivers and omission of services for the elderly. The structured output of the allocation matrix is also convenient for direct integration into the scheduling system, forming a closed-loop automated task management process, with strong system compatibility and practicality.

[0157] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0158] The above are only the preferred embodiments of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. An automatic generation and allocation method for elderly care tasks, characterized in that, Including: Collect the basic data of the elderly and caregivers, and define the original variables, including the daily behavior characteristics of the elderly and the service ability parameters of the caregivers; Standardize the original feature data of the elderly to form a standardized feature vector; Calculate the feature distance between any two elderly based on the standardized features; Calculate the distance scale parameter according to the feature distance and perform adaptive adjustment; Construct a similarity matrix based on the distance scale parameter and calculate the similarity threshold; Use the similarity matrix to construct an undirected graph and extract the connected components; Select a subset of caregivers that meet the needs of the elderly based on the greedy strategy; Allocate care tasks according to the ratio of the needs of the elderly and the abilities of the caregivers, and output the task allocation results.

2. The automatic generation and allocation method of elderly care tasks according to claim 1, wherein The collection of the basic data of the elderly and caregivers and the definition of the original variables specifically include: Let the total number of the elderly be , and the elderly index set be . The index corresponds to the th elderly person, denoted as ; Let the total number of caregivers be , and the caregiver index set be , the index corresponds to the th caregiver, denoted as ; For each elderly person Collect five-dimensional original features , including: , average daily number of steps walked in the past 7 days; , average daily sleep duration in the past 7 days; , average daily number of toilet visits in the past 7 days; , average daily number of meals in the past 7 days; , average daily number of times of taking medicine in the past 7 days; among them, is the original feature index; Suppose the elderly The duration of the care needs on that day is , and the caregiver The available service duration on that day is .

3. The automatic generation and allocation method of elderly care tasks according to claim 2, characterized in that The standardization process of the original feature data of the elderly to form a standardized feature vector specifically includes: Calculate the arithmetic mean of the original features of the -dimensionality; Calculate the standard deviation of the original features of the -dimensional; Constructed elderly of the standardized eigenvector , ; among which, is the -dimensional standardized value.

4. The automatic generation and allocation method of elderly care tasks according to claim 3, characterized in that The calculation of the feature distance between any two elderly based on the standardized features specifically includes: For any two elderly indexes , calculate the Euclidean distance between them : 。 5. The automatic generation and allocation method of elderly care tasks according to claim 4, characterized in that The calculation of the distance scale parameter according to the feature distance and the adaptive adjustment specifically include: Set the overall distance scale parameter to .

6. The automatic generation and allocation method of elderly care tasks according to claim 5, characterized in that The construction of the similarity matrix based on the distance scale parameter and the calculation of the similarity threshold specifically include: Set the elderly and The demand similarity is , ; The construction similarity set is ; Calculate the median threshold ; wherein is the median function, which takes the median value after sorting .

7. The automatic generation and allocation method of elderly care tasks according to claim 6, characterized in that, The construction of the undirected graph using the similarity matrix and the extraction of the connected components specifically include: Construct an adjacency indicator function: ; wherein, is an adjacency indication variable, which takes the value of 1 if and only if the similarity between the two elderly people is not lower than the threshold [[ID=.. and represents that an edge should be connected between vertices and in the undirected graph; otherwise it is 0 and no edge is connected; Initialize the set of unvisited vertex indices , the connected component counter ; When Repeat the following steps: S1100. Let ; S1200. Take out the minimum index from , and set the current component vertex set . Then delete from . S1300. Set pointer set ; S1400. When : S1410. Optionally , delete it from ; S1420, for all Check, if , then execute: S1421. Add to ; S1422. Add to ; S1423. Delete from ; ; S1500, until when it is the vertex set of the -th connected component, and denote its edge set as: ； S1600. Repeat steps S1100 to S1500 until ; Finally, a set of connected components is obtained .

8. A method for automatically generating and allocating elderly care tasks according to claim 7, characterized in that, The selection of a subset of caregivers that meet the needs of the elderly based on the greedy strategy specifically includes: For each component Perform the following steps: S201. Calculate the total care needs of all the elderly in ; S202. Sort the indexes in from largest to smallest, and denote the sorted sequence as , satisfying ; where ; among them, is the index of the th largest available duration caregiver; S203. Let calculate the cumulative available duration ; where is a counter used to record the number of members currently selected into the caregiver subset If ,let Repeat; otherwise stop; S204. Select the minimum set of caregivers that meets the component demand , ; where is the total available duration of the set .

9. The automatic generation and allocation method of elderly care tasks according to claim 8, characterized in that The allocation of care tasks according to the ratio of the needs of the elderly and the abilities of the caregivers and the output of the task allocation results specifically include: For each and each caregiver calculate: ; where is the component of the elderly assigned to the caregiver service duration; Verification: S301, Satisfaction of the elderly's needs: ; S302. Non-overloading of caregivers: ; For each component Output: List of the elderly population subgroups , subset of caregivers and allocation matrix .

Images