An old-age service quality intelligent evaluation method and system based on multi-dimensional data fusion

Abstract

The application provides a pension service quality intelligent evaluation method and system based on multi-dimensional data fusion, and relates to the technical field of pension service data analysis. The method comprises the following steps: collecting historical pension service record data of a plurality of old people, generating a global service preference matrix and a local service fluctuation matrix of each old person; constructing a service correlation graph of each old person, calculating the interaction influence parameter between any two pension services and generating a historical interaction influence matrix; encoding the service trajectory of the historical pension service record data, and constructing a micro-service mode set of each old person; calculating a plurality of individual behavior linkage weights of each old person according to the micro-service mode set, calculating the global preference parameter of the old people on each pension service, and performing pension service quality evaluation to generate the pension service quality evaluation result of each old person. The application realizes accurate pension service quality evaluation based on individual demand characteristics.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pension service data analysis, and in particular to a pension service quality intelligent evaluation method and system based on multi-dimensional data fusion. BACKGROUND

[0002] Intelligent evaluation of pension service quality is an important part of pension service system construction. Some pension service quality evaluation methods use unified standards or expert scoring methods to comprehensively score different types of pension service projects such as medical care, social entertainment, and psychological counseling. However, in the actual pension service process, different elderly people have significant differences in their attention, sensitivity, and actual experience of the same service project due to individual needs and behavioral characteristics. For example, some elderly people may pay more attention to medical care services, while others may focus more on social activities or psychological support. Moreover, pension service needs are not isolated, and there are complex complementary, alternative, and linkage relationships between different service projects, such as an increase in medical care needs often accompanied by a decrease in social activities, and the use of intelligent auxiliary equipment may reduce the demand for human care.

[0003] If a single scoring weight is obtained by ignoring individual differences, the same set of service evaluation indicators and weights is applied to all elderly people, which may result in a low reference value of the evaluation results of the pension service quality due to the lack of in-depth analysis of individual preferences and behavioral differences of the elderly, and may not achieve truly individualized service quality evaluation centered on the elderly. SUMMARY

[0004] To solve the above technical problems, the present application provides a pension service quality intelligent evaluation method and system based on multi-dimensional data fusion, which dynamically generates individualized inclination weights of each elderly person for different pension service projects through service association network modeling and service behavior chain analysis, and can better achieve accurate pension service quality evaluation based on individual demand characteristics.

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] As a first aspect of the present application, a pension service quality intelligent evaluation method based on multi-dimensional data fusion is provided, comprising:

[0007] Collecting historical pension service record data of a plurality of elderly people, constructing a use feature sequence of each elderly person for each pension service and performing window decomposition to generate a global service preference matrix and a local service fluctuation matrix of each elderly person;

[0008] According to the interaction influence analysis of a plurality of service features on a plurality of pension services, a service correlation graph of each elderly person is constructed, and an interaction influence parameter between any two pension services is calculated according to a plurality of service correlation graphs to generate a historical interaction influence matrix of the plurality of pension services;

[0009] The service trajectory of the historical pension service record data is encoded, the pension service link of each elderly person is constructed, the service chain of each pension service link is segmented to generate a plurality of candidate service chains, and a micro-service mode set of each elderly person containing a plurality of target service chains is generated.

[0010] The micro-service mode set of each elderly person is interactively processed through the historical interaction influence matrix, and an individual behavior linkage weight corresponding to each target service chain of each elderly person is calculated.

[0011] According to the individual behavior linkage weight, the global service preference matrix and the local service fluctuation matrix, a global preference parameter of each elderly person on each pension service is generated, and a pension service quality evaluation result of each elderly person is generated according to the global preference parameter.

[0012] Preferably, the micro-service mode set of each elderly person containing a plurality of target service chains is generated, comprising:

[0013] A plurality of service behaviors of the elderly person in the historical pension service record data are determined, and a pension service link of the elderly person on a plurality of pension services is constructed based on the time sequence according to the time stamp and the corresponding service item of the service behavior.

[0014] The service chain of the pension service link is segmented to generate a plurality of candidate service chains based on a preset sequence length, the distribution frequency of each candidate service chain under the pension service link is counted, and the mode distribution sequence of each candidate service chain is constructed according to a plurality of time stamps of the candidate service chain. The behavior regularity of each candidate service chain is analyzed according to the mode distribution sequence, the behavior regularity index of each candidate service chain is calculated, the target service chain is determined from the plurality of candidate service chains according to the behavior regularity index, and the micro-service mode set of each elderly person is constructed.

[0015] Preferably, the micro-service mode set of each elderly person is interactively processed through the historical interaction influence matrix, and an individual behavior linkage weight corresponding to each target service chain of each elderly person is calculated, comprising:

[0016] The starting point and the ending point of the target service chain are determined to determine the pension service combination corresponding to each target service chain, the interaction influence parameter of each target service chain is extracted from the historical interaction influence matrix according to the pension service combination, and the individual behavior linkage weight of the target service chain is calculated based on the distribution frequency and the interaction influence parameter of the target service chain.

[0017] Preferably, the global preference parameter of the elderly person on each pension service is generated according to the individual behavior linkage weight, the global service preference matrix and the local service fluctuation matrix, including:

[0018] The historical interaction influence matrix is corrected according to the individual behavior linkage weight of the plurality of target service chains to generate a dynamic interaction influence matrix, and the micro-service mode set of each elderly person is interactively processed according to the dynamic interaction influence matrix to calculate the target behavior linkage weight corresponding to each target service chain of the plurality of target service chains of each elderly person;

[0019] The plurality of target service chains associated with each pension service of the elderly person is determined, the chain tendency factor of the elderly person on each pension service is generated according to the target behavior linkage weight of the target service chain, the long-term and short-term preference tendency factor of each pension service of the elderly person is calculated according to the global service preference matrix and the local service fluctuation matrix, and the global preference parameter of the elderly person on each pension service is generated according to the chain tendency factor and the long-term and short-term preference tendency factor.

[0020] Preferably, for the historical interaction influence matrix, characterized in that,

[0021] The initial correlation coefficient between the use feature sequences of any two pension services is calculated, the service correlation graph of the elderly person is constructed according to the initial correlation coefficient and the preset correlation threshold, the correlation degree number of each elderly person on each pension service is determined according to the service correlation graph, the initial correlation coefficient between any two pension services of the elderly person is corrected based on the correlation degree number of the pension service to generate the target correlation coefficient between any two pension services, and the local interaction matrix of each elderly person is constructed;

[0022] The plurality of local interaction matrices are fused, including determining the plurality of target correlation coefficients corresponding to any two pension services according to the plurality of local interaction matrices, calculating the interaction influence parameters between any two pension services, and generating the historical interaction influence matrix according to the plurality of interaction influence parameters.

[0023] Preferably, the behavior regularity analysis is performed on each candidate service chain according to the mode distribution sequence to calculate the behavior regularity index of each candidate service chain, including:

[0024] The time interval between any adjacent two candidate service chains in the mode distribution sequence is counted, the standard deviation of the plurality of time intervals is calculated, and the ratio between the distribution frequency of the candidate service chain and the standard deviation of the plurality of time intervals is taken as the behavior regularity index.

[0025] As a second aspect of the present application, a pension service quality intelligent evaluation system based on multi-dimensional data fusion is provided for implementing the pension service quality intelligent evaluation method based on multi-dimensional data fusion described above, comprising:

[0026] a service use feature analysis module for collecting historical pension service record data of a plurality of old people, constructing a use feature sequence of each old person about each pension service and performing window decomposition, generating a global service preference matrix and a local service fluctuation matrix of each old person;

[0027] a service interaction influence analysis module for performing interaction influence analysis on a plurality of pension services according to a plurality of use feature sequences, constructing a service correlation graph of each old person, calculating an interaction influence parameter between any two pension services according to a plurality of service correlation graphs, and generating a historical interaction influence matrix about a plurality of pension services;

[0028] a micro-service link analysis module for performing service trajectory coding on historical pension service record data, constructing a pension service link of each old person, performing service chain segmentation on each pension service link to generate a plurality of candidate service chains, and generating a micro-service mode set of each old person containing a plurality of target service chains;

[0029] an individual behavior linkage analysis module for performing interaction processing on the micro-service mode set of each old person through the historical interaction influence matrix respectively, and calculating individual behavior linkage weights corresponding to a plurality of target service chains of each old person respectively;

[0030] a pension service quality evaluation management module for generating a global preference parameter of each old person about each pension service according to the individual behavior linkage weight, the global service preference matrix and the local service fluctuation matrix, performing pension service quality evaluation according to the global preference parameter, and generating a pension service quality evaluation result of each old person.

[0031] The present application has the following beneficial effects:

[0032] The application analyzes the historical pension service record data of the elderly by dynamics, mines the long-term and short-term service demand dynamic change relationship of the elderly about pension services, generates the global service preference matrix and the local service fluctuation matrix of each elderly person, analyzes the interactive correlation characteristics between different pension services, obtains the historical interactive influence matrix about multiple pension services, and further analyzes the individual behavior chain of each elderly person, extracts through the micro-service mode set, adjusts the correlation between services in combination with the dynamic interactive influence matrix, dynamically identifies the individualization tendency of each elderly person to different pension service projects, generates the individualized service weight conforming to the actual demand characteristics of the elderly, realizes the individualized and accurate evaluation of the pension service quality, and makes the pension service quality evaluation result more conform to the actual demand and feeling of the elderly, and improves the intelligent level of the pension service quality evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A flowchart of a pension service quality intelligent evaluation method based on multi-dimensional data fusion provided by the embodiment of the application is shown.

[0034] Figure 2 A structure diagram of a pension service quality intelligent evaluation system based on multi-dimensional data fusion provided by the embodiment of the application is shown. DETAILED DESCRIPTION

[0035] In order for those skilled in the art to better understand the technical solutions in the application, the technical solutions in the embodiments of the application will be described clearly and completely in combination with the drawings in the embodiments of the application. It should be understood that the specific embodiments described herein are only used to explain the application and not to limit the application.

[0036] Please refer to Figure 1 The pension service quality intelligent evaluation method based on multi-dimensional data fusion provided by the embodiment of the application includes the following steps.

[0037] Step S1, collect the historical pension service record data of a plurality of elderly people, construct the use feature sequence of each elderly person about each pension service and perform window decomposition, and generate the global service preference matrix and the local service fluctuation matrix of each elderly person.

[0038] Specifically, the service record data of a plurality of old people in a period of time is collected, and the collected data includes but is not limited to the use time, use frequency, service type and the like of various pension services, such as service record data of the old people in medical care, social entertainment, psychological counseling and the like. For each old person, for all pension service projects, a use feature sequence corresponding to each pension service is constructed in chronological order, and a window decomposition technique is used to divide each use feature sequence into a plurality of time windows, and the mean feature of the use frequency of each pension service in each window is extracted, representing a long-term stable preference of the old people in a specific period. At the same time, the fluctuation feature of each window about each pension service, such as the standard deviation of the use frequency, is extracted, representing the short-term fluctuation of the old people in a short period due to special events such as illness, etc. The selection tendency of the pension service. Thus, a global service preference matrix is generated for depicting the overall service preference and long-term dependence relationship of the old people, and a local service fluctuation matrix is constructed for depicting the volatility and sensitivity of the service demand of the old people in a short period.

[0039] Step S2, according to a plurality of use feature sequences, the interaction influence analysis of a plurality of pension services is carried out, the service association graph of each old person is constructed, the interaction influence parameter between any two pension services is calculated according to a plurality of service association graphs, and the historical interaction influence matrix of a plurality of pension services is generated.

[0040] Specifically, for a plurality of use feature sequences of each old person, the linkage relationship between pension service projects is analyzed in depth, the interaction influence features of different service projects in use behavior are identified, the correlation between the use frequencies of different pension services is analyzed, and the service association graph of each old person is constructed. And based on a plurality of inspiration association graphs, global analysis is carried out, the overall interaction influence features between different pension services are evaluated from the overall point of view, the interaction influence parameters between any two pension services are calculated, and finally the historical interaction influence matrix corresponding to a plurality of pension services is constructed.

[0041] In an alternative embodiment, for the construction of the historical interaction influence matrix, the initial correlation coefficient between the use feature sequences of any two pension services is calculated first. In this embodiment, the mutual information between the two sequences is taken as the initial correlation parameter. The service correlation graph of the elderly is constructed according to the initial correlation coefficient and the preset correlation threshold. The service correlation graph takes the pension service as the node and the interaction correlation between the pension services as the edge. The edge weight represents the correlation interaction strength between the service items, that is, the initial correlation coefficient. In order to improve the representativeness of the service correlation graph, only when the initial correlation parameter of the two services is higher than the preset correlation threshold, it is considered that there is an effective correlation between the two services, so as to form a dynamic service relationship network reflecting the service behavior characteristics of the elderly individuals. According to the service correlation graph, the correlation degree of each elderly person with respect to each pension service is determined, which is the total number of graph edges of each pension service. The higher the number of retained edges, the higher the overall importance of the service.

[0042] In order to avoid the problem of high or low initial correlation coefficient caused by some high-frequency but low-dependence services, the initial correlation coefficient between any two pension services of the elderly is corrected in combination with the correlation degree of the pension services. For example, after normalizing the overall correlation degree, the correlation degrees corresponding to the two pension services are added as the correction weight. The initial correlation coefficient with high correlation degree combination is retained as much as possible, and the coefficient with low correlation degree combination is reduced and weakened, so as to generate the target correlation coefficient between any two pension services, and to construct the local interaction matrix of each elderly person.

[0043] The local interaction matrices corresponding to multiple elderly persons are fused. For the multiple target correlation coefficients corresponding to any two pension services, in this embodiment, the sum and the average of the target correlation coefficients are taken as the interaction influence parameter between the two pension services, so as to realize the fusion of multiple local interaction matrices, and to construct the historical interaction influence matrix representing the overall correlation characteristics of different pension services, reflecting the comprehensive interaction relationship between all pension services under the actual behavior background of the elderly.

[0044] Step S3, service trajectory coding is performed on the historical pension service record data, and the pension service link of each elderly person is constructed. The service chain of each pension service link is segmented to generate multiple candidate service chains, and the micro-service mode set of each elderly person containing multiple target service chains is generated.

[0045] Specifically, for multiple sets of historical pension service record data, the pension service usage trajectory of each elderly person is encoded to form a complete pension service link. The link reflects the process of the elderly person using different pension service items in sequence in a time series manner, and reveals the inherent logic of the service combination and behavior path of the elderly person. Based on the pension service link, chain segmentation is performed to segment the complete behavior link to generate multiple candidate service chains, and representative candidate service chains, for example, candidate service chains with higher frequency and stronger stability in the pension service link, are extracted to form a micro-service mode set of each elderly person. The set can reflect the joint demand and behavior habit of the elderly person for multiple services, and reflect the real use relationship and behavior chain characteristics between services.

[0046] In an optional embodiment, for the construction process of the micro-service mode set of each elderly person containing multiple target service chains, multiple service behaviors of the elderly person in the historical pension service record data are determined, and based on the time stamp and corresponding service item of the service behavior, a pension service link of the elderly person about multiple pension services is constructed based on the time sequence. Reflects the real path and process of the elderly person using pension services.

[0047] For the complete pension service link, service chain segmentation is performed on the pension service link based on a preset sequence length to generate multiple candidate service chains. The preset sequence length can be, for example, a combination of 2, 3, or 4 consecutive services. In this embodiment, two consecutive pension service combinations are taken as an example, and a sliding window is used to extract multiple micro-chains of fixed length in order to obtain multiple candidate service chains of each elderly person. The distribution frequency of each candidate service chain under the pension service link is further counted, and a mode distribution sequence of each candidate service chain is constructed according to the multiple time stamps of the candidate service chain. Then, according to the mode distribution sequence, the behavior regularity analysis of each candidate service chain is performed, and the behavior regularity index of each candidate service chain is calculated. For example, the time interval between any two adjacent candidate service chains in the mode distribution sequence is counted, the standard deviation of the multiple time intervals is calculated, and the ratio between the distribution frequency of the candidate service chain and the standard deviation of the multiple time intervals is taken as the behavior regularity index, which is used to comprehensively reflect the behavior stability and regularity. Finally, according to the behavior regularity index, multiple representative target service chains are determined from the multiple candidate service chains, and a micro-service mode set of each elderly person is constructed to comprehensively and meticulously reflect the stable and preferred service link of the elderly person in actual service behavior.

[0048] Step S4, the micro-service mode set of each elderly person is interactively processed by the historical interaction influence matrix, and the individual behavior linkage weight corresponding to each target service chain of each elderly person is calculated.

[0049] Specifically, on the basis of the aforementioned extracted micro-service mode set, in combination with the historical interaction influence matrix, the interaction weight dynamic coupling is performed for each micro-service chain of each elderly person, the individual behavior linkage weight is formed by comprehensively considering the individual behavior chain frequency and the historical global level service interaction relationship.

[0050] In the embodiment, the starting point and the ending point of the target service chain are determined to determine the pension service combination corresponding to each target service chain, that is, the pension services corresponding to the starting point and the ending point of the target service chain, the interaction influence parameters corresponding to the reorganized combination are extracted from the historical interaction influence matrix to serve as the interaction influence parameters of each target service chain, and finally the individual behavior linkage weight of the target service chain is calculated based on the distribution frequency and the interaction influence parameters of the target service chain. That is, for each target service chain of each elderly person, the relative importance and frequency of the service chain in the individual behavior of the elderly are comprehensively considered according to the interaction influence degree between the service items in the chain, for example, the fusion weight is determined according to the distribution frequency, and the interaction influence parameters are modified and fused to obtain the individual behavior linkage weight of the target service chain. Through this dynamic fusion mode of behavior chain and service interaction, the service linkage mode existing in the actual life of the elderly can be better represented, the interference of single service behavior or accidental behavior on the overall model is avoided, and the real service use logic of the elderly is more accurately captured.

[0051] In step S5, the global preference parameters of each pension service of the elderly are generated according to the individual behavior linkage weight, the global service preference matrix and the local service fluctuation matrix, the pension service quality evaluation is performed according to the global preference parameters, and the pension service quality evaluation result of each elderly person is generated.

[0052] Specifically, the individual behavior linkage weight obtained in the foregoing steps, and the global service preference matrix and the local service fluctuation matrix are used for multi-dimensional fusion analysis of the individualized tendency of each elderly person on each pension service, and the global preference parameters reflecting the individual service demand are generated, so as to comprehensively reflect the multi-dimensional factors such as long-term dependence, recent demand fluctuation and service linkage of the elderly, and effectively depict the real sensitivity and preference weight of the elderly to different pension service items.

[0053] Finally, according to the global preference parameters, in combination with the preset quality score standards of each pension service, for example, the scores of different pension services obtained through expert evaluation, the service quality evaluation of all pension services based on the individualized weight is performed, and the individualized pension service quality evaluation result of each elderly person is generated. Compared with the evaluation method of part of single weight, the evaluation weight of each service can be dynamically adjusted according to the actual service demand of each elderly person, so that more accurate intelligent evaluation of the pension service quality is realized, and the individualization and dynamic level of the pension service management is improved.

[0054] In an alternative embodiment, the global preference parameter of the elderly for each pension service is generated according to the individual behavior linkage weight, the global service preference matrix and the local service fluctuation matrix, including:

[0055] The historical interaction influence matrix is corrected according to the individual behavior linkage weight of the plurality of target service chains to generate a dynamic interaction influence matrix of the historical interaction influence matrix. In this process, for any one pension service combination in the historical interaction influence matrix, on the basis of the interaction influence parameter of the pension service combination, the individual behavior linkage weight of the plurality of target service chains involving the pension service combination is calculated as a correction weight after averaging, and the interaction influence parameter of the pension service combination is dynamically adjusted to generate a dynamic interaction influence matrix with individual characteristics to reflect the influence of individual behavior characteristics on the interaction relationship between services.

[0056] According to the dynamic interaction influence matrix, the micro-service mode set of each elderly person is interactively processed to calculate the target behavior linkage weight corresponding to each target service chain of each elderly person. That is, the calculation method of the individual behavior linkage weight is the same, and the dynamically adjusted dynamic interaction influence matrix is replaced by the original historical interaction influence matrix based on the historical overall law.

[0057] For the plurality of target service chains associated with each pension service of the elderly, which can be the target service chain with the pension service as the starting point or the end point, is recorded as having relevance. According to the target behavior linkage weight corresponding to the plurality of target service chains having relevance, the chain tendency factor of the elderly for each pension service is generated after summation, reflecting the actual behavior importance of the pension service as the core node of the chain, which can accurately reflect the comprehensive behavior performance of the elderly in the actual life, the dependence and linkage characteristics of a pension service and other services.

[0058] Further according to the global service preference matrix and the local service fluctuation matrix, a long-term and short-term preference tendency factor of each pension service of the old people is calculated. In this process, considering that the selection tendency of the old people on the pension service may change in different periods, the long-term and short-term preference tendency factor in a specific period can be extracted according to the global service preference matrix and the local service fluctuation matrix, for example, the ratio of the average use frequency of a certain pension service to the standard deviation in a specific time is taken as the long-term and short-term preference tendency factor in the period, and according to the chain tendency factor and the long-term and short-term preference tendency factor, a global preference parameter of the old people on the pension service in a specific period is obtained after weighted fusion, and the weight of the weighted fusion can be set according to the overall proportion of the chain tendency factor and the long-term and short-term preference tendency factor, or the weight of the chain tendency factor and the long-term and short-term preference tendency factor can be quantified based on expert experience, and the importance of the pension service between the combination behavior and the long-term dependence behavior is reasonably set.

[0059] It is worth noting that the demand of each old person for different pension services may have different degrees of individual differences, and some pension service quality evaluation methods with the same evaluation weight for each pension service are difficult to reflect individual differences and may only have high matching degree for part of the old people, while some groups with individual demand different from group demand cannot well evaluate the matching degree of the current pension service, which may lead to the final pension service evaluation result not representative in the case of large individual preference difference. The global preference parameter constructed by the present application comprehensively considers the service dependence and linkage relationship of the old people in the behavior chain, and the long-term rigid demand and short-term temporary demand for the service, can dynamically and individually depict the actual attention of different old people for each pension service, embodies the people-centered service thinking, and ensures that the evaluation result is closer to the real feelings of the old people. For example, the weight of medical service is high for the old people sensitive to medical service, and the weight of social service is lower for the old people not interested in social service, and the final comprehensive evaluation result is more in line with the actual experience, providing a good data basis for constructing a personalized pension service system.

[0060] Please refer to Figure 2 The pension service quality intelligent evaluation system based on multi-dimensional data fusion provided by the embodiment of the present application is specifically based on the concept of the above-mentioned pension service quality intelligent evaluation method based on multi-dimensional data fusion, and comprises:

[0061] The service use feature analysis module is used for collecting historical pension service record data of a plurality of old people, constructing a use feature sequence of each old person on each pension service and performing window decomposition, and generating a global service preference matrix and a local service fluctuation matrix of each old person.

[0062] The service interaction influence analysis module is configured to perform interaction influence analysis on multiple pension services according to a plurality of use feature sequences, construct a service correlation graph for each elderly person, calculate an interaction influence parameter between any two pension services according to a plurality of service correlation graphs, and generate a historical interaction influence matrix of the multiple pension services.

[0063] The micro-service link analysis module is configured to perform service trajectory coding on the historical pension service record data, construct a pension service link for each elderly person, perform service chain segmentation on each pension service link to generate a plurality of candidate service chains, and generate a micro-service mode set of each elderly person containing a plurality of target service chains.

[0064] The individual behavior linkage analysis module is configured to perform interaction processing on the micro-service mode set of each elderly person through the historical interaction influence matrix, and calculate individual behavior linkage weights corresponding to the plurality of target service chains of each elderly person.

[0065] The pension service quality evaluation management module is configured to generate a global preference parameter of each elderly person for each pension service according to the individual behavior linkage weights, a global service preference matrix, and a local service fluctuation matrix, perform pension service quality evaluation according to the global preference parameter, and generate a pension service quality evaluation result of each elderly person.

[0066] The above is only a specific embodiment of the present application, enabling those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art. The parts not described in detail in the specification belong to the prior art known to those skilled in the art.

Claims

1. An old-age service quality intelligent evaluation method based on multi-dimensional data fusion, characterized in that, Comprise: Collect historical pension service record data of a plurality of elderly people, construct the use feature sequence of each elderly person about each pension service and perform window decomposition, generate the global service preference matrix and the local service fluctuation matrix of each elderly person; According to a plurality of use feature sequences, the interaction influence analysis is carried out on a plurality of pension services, the service association graph of each elderly person is constructed, the interaction influence parameter between any two pension services is calculated according to a plurality of service association graphs, and the historical interaction influence matrix about a plurality of pension services is generated; Service trajectory coding is performed on the historical pension service record data, the pension service link of each elderly person is constructed, the service chain segmentation is performed on each pension service link to generate a plurality of candidate service chains, and the micro service mode set of each elderly person containing a plurality of target service chains is generated, including determining a plurality of service behaviors of the elderly people in the historical pension service record data, constructing the pension service link of the elderly people about a plurality of pension services based on the time sequence according to the time stamp and the corresponding service item of the service behavior; Based on the preset sequence length, the service chain segmentation is performed on the pension service link to generate a plurality of candidate service chains, the distribution frequency of each candidate service chain under the pension service link is counted, and the mode distribution sequence of each candidate service chain is constructed according to a plurality of time stamps of the candidate service chain, the behavior regularity analysis is performed on each candidate service chain according to the mode distribution sequence, the behavior regularity index of each candidate service chain is calculated, the plurality of target service chains are determined from the plurality of candidate service chains according to the behavior regularity index, and the micro service mode set of each elderly person is constructed; The micro service mode set of each elderly person is respectively interacted through the historical interaction influence matrix, the individual behavior linkage weight corresponding to each target service chain of the elderly person is calculated, including determining the pension service combination corresponding to each target service chain according to the starting point and the ending point of the target service chain, extracting the interaction influence parameter of each target service chain from the historical interaction influence matrix according to the pension service combination, and calculating the individual behavior linkage weight of the target service chain based on the distribution frequency and the interaction influence parameter of the target service chain; According to the individual behavior linkage weight, the global service preference matrix and the local service fluctuation matrix, the global preference parameter of the elderly people about each pension service is generated, including correcting the historical interaction influence matrix according to the individual behavior linkage weight of a plurality of target service chains to generate a dynamic interaction influence matrix of the historical interaction influence matrix, and interactively processing the micro service mode set of each elderly person according to the dynamic interaction influence matrix to calculate the target behavior linkage weight corresponding to each target service chain of the elderly person; The method comprises the following steps: determining a plurality of target service chains associated with each pension service of the elderly, generating a chain tendency factor of the elderly for each pension service according to a target behavior linkage weight value of the target service chain, calculating a long-term and short-term preference tendency factor of the elderly for each pension service according to a global service preference matrix and a local service fluctuation matrix, generating a global preference parameter of the elderly for each pension service according to the chain tendency factor and the long-term and short-term preference tendency factor, and performing pension service quality evaluation according to the global preference parameter to generate a pension service quality evaluation result of each elderly person.

2. The method according to claim 1, wherein, for the historical interaction influence matrix, the method comprises the following steps: calculating an initial correlation coefficient between the use characteristic sequences of any two pension services, constructing a service correlation graph of the elderly according to the initial correlation coefficient and a preset correlation threshold, determining a correlation degree of each elderly person for each pension service according to the service correlation graph, modifying the initial correlation coefficient between any two pension services of the elderly according to the correlation degree of the pension services to generate a target correlation coefficient between any two pension services, and constructing a local interaction matrix of each elderly person. The method comprises the following steps: calculating an initial correlation coefficient between the use characteristic sequences of any two pension services, constructing a service correlation graph of the elderly according to the initial correlation coefficient and a preset correlation threshold, determining a correlation degree of each elderly person for each pension service according to the service correlation graph, modifying the initial correlation coefficient between any two pension services of the elderly according to the correlation degree of the pension services to generate a target correlation coefficient between any two pension services, and constructing a local interaction matrix of each elderly person. The method comprises the following steps: calculating an initial correlation coefficient between the use characteristic sequences of any two pension services, constructing a service correlation graph of the elderly according to the initial correlation coefficient and a preset correlation threshold, determining a correlation degree of each elderly person for each pension service according to the service correlation graph, modifying the initial correlation coefficient between any two pension services of the elderly according to the correlation degree of the pension services to generate a target correlation coefficient between any two pension services, and constructing a local interaction matrix of each elderly person. 3.The method of claim 2, wherein, The system is used to implement the method according to any one of claims 1-3, and comprises: a service use characteristic analysis module configured to collect historical pension service record data of a plurality of elderly persons, construct a use characteristic sequence of each elderly person for each pension service, and perform window decomposition to generate a global service preference matrix and a local service fluctuation matrix of each elderly person; 4. An old-age service quality intelligent evaluation system based on multi-dimensional data fusion, characterized in that, a service interaction influence analysis module configured to perform interaction influence analysis on a plurality of pension services according to a plurality of use characteristic sequences, construct a service correlation graph of each elderly person, calculate an interaction influence parameter between any two pension services according to a plurality of service correlation graphs, and generate a historical interaction influence matrix of a plurality of pension services; a micro-service link analysis module configured to perform service trajectory coding on the historical pension service record data, construct a pension service link of each elderly person, perform service chain segmentation on each pension service link to generate a plurality of candidate service chains, and generate a micro-service mode set of each elderly person comprising a plurality of target service chains. ​ ​ An individual behavior linkage analysis module is configured to respectively interact with the micro-service mode set of each elderly person through a historical interaction influence matrix, and calculate individual behavior linkage weight values corresponding to a plurality of target service chains of each elderly person. An old-age service quality evaluation management module is configured to generate a global preference parameter of each elderly person on each old-age service according to the individual behavior linkage weight values, a global service preference matrix and a local service fluctuation matrix, perform old-age service quality evaluation according to the global preference parameter, and generate an old-age service quality evaluation result of each elderly person.

Images