JP2015052859A - Service analysis device and its operation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a service analysis device which performs an analysis for each execution unit for service that a plurality of operators perform work for a plurality of objects.SOLUTION: In the service analysis device, a work amount calculation section calculates work amounts for each of a plurality of objects on the basis of service history data. A clustering section generates management information indicating a degree that each of a plurality of execution units manages each of the plurality of objects and affiliation information indicating a degree that each of a plurality of operators belong to the plurality of execution units on the basis of the service history data. An incident amount calculation section calculates an incident amount for each of the plurality of objects on the basis of the incident data. A state analysis section analyses a work state and an incident state for each of the plurality of execution units on the basis of the work amount, the affiliation information, the management information, and the incident amount.

Description

  Embodiments described herein relate generally to a service analysis apparatus and method.

  Business support systems that use computing devices have been introduced in services centered on people, such as equipment maintenance services, medical services, and nursing care services. As a result, service history data is accumulated in an organization that provides services using the business support system. In such an organization, the service is improved by using an analyzer that evaluates service quality by analyzing service history data.

JP 2006-146735 A JP 2007-241331 A

  In a conventional analyzer, analysis is performed for each object (project) or worker (individual). However, when a plurality of workers work together like a maintenance service of equipment (for example, an elevator etc.), it is more accurate to perform analysis for each group (working unit) that performed the work. . Although it is possible to identify the implementation unit using the organization database, the current situation at the site is not reflected well in the analysis using the organization database as it is because of support beyond the organization or the existence of subunits within the organization. There is a problem.

  The problem to be solved by the present invention is to provide a service analysis apparatus and method capable of analyzing for each execution unit a service in which a plurality of workers work on a plurality of objects. .

  A service analysis apparatus according to an embodiment evaluates the quality of a service in which a plurality of workers perform work on a plurality of objects, and includes a work amount calculation unit, a clustering unit, an incident amount calculation unit, and A situation analysis unit is provided. The work amount calculation unit calculates the work amount for each of the plurality of objects based on service history data related to the performed work. The clustering unit, based on the service history data, management information indicating a degree to which each of the plurality of execution units that group the plurality of workers manages the plurality of objects, and the plurality of workers And affiliation information indicating the degree to which each of the plurality of execution units belongs. The incident amount calculation unit is an operation for dealing with an incident that has occurred in the plurality of objects, and calculates an incident amount for each of the plurality of objects based on incident data related to the performed work. The situation analysis unit, based on the work amount for each of the plurality of objects, the affiliation information, the management information, and the incident amount for each of the plurality of objects, Analyze the incident status for each execution unit.

1 is a block diagram schematically showing a service analysis apparatus according to an embodiment. The figure which shows an example of the service history database shown in FIG. The figure which shows an example of the operator database shown in FIG. The figure which shows an example of the service object database shown in FIG. The figure which shows an example of the incident database shown in FIG. The figure which shows an example of the work amount computed for every target object by the work amount calculation part shown in FIG. The figure which shows an example of the work matrix computed by the work matrix calculation part shown in FIG. FIG. 8 is a diagram illustrating a result of co-clustering performed on the work matrix in FIG. 7 by the clustering unit illustrated in FIG. 1. The figure which shows an example of the affiliation information produced | generated by the clustering part shown in FIG. The figure which shows an example of the management information produced | generated by the clustering part shown in FIG. The figure which shows an example of the incident amount calculated for every target object by the incident amount calculation part shown in FIG. The figure which shows the analysis result by the analysis part shown in FIG. (A) is a scatter diagram showing the correlation between the workload and the incident occurrence rate, and (b) is a scatter diagram showing the correlation between the work amount per object and the incident occurrence rate. The flowchart which shows an example of the analysis process procedure of the service analysis apparatus shown in FIG. The figure explaining the co-clustering which the clustering part shown in FIG. 1 performs. The figure explaining the normalization process of the normalization part shown in FIG. The flowchart which shows an example of the normalization process procedure of the normalization part shown in FIG. The figure which shows the other example of a work matrix, and the management matrix and affiliation matrix produced | generated from this work matrix. The block diagram which shows roughly the hardware constitutions of the service analysis apparatus shown in FIG.

Hereinafter, embodiments will be described with reference to the drawings.
FIG. 1 schematically shows a service analysis apparatus 100 according to the embodiment. The service analysis apparatus 100 analyzes data related to a service in which a plurality of workers perform work on a plurality of objects, and evaluates service quality for each execution unit. The execution unit indicates a group of workers who frequently work on the same object. Each worker can belong to one or more implementation units. In the present embodiment, an elevator maintenance service is assumed as an example of a service in which a plurality of workers perform work on a plurality of objects.

  As shown in FIG. 1, the service analysis apparatus 100 includes a work amount calculation unit 101, a clustering unit 102, an incident amount calculation unit 106, an affiliation information storage unit 107, a management information storage unit 108, a situation analysis unit 109, and an analysis result output unit. 113 and an input unit 114. A service history database 151, a worker database 152, a service target database 153, and an incident database 154 are prepared as databases for storing data related to the elevator maintenance service. Hereinafter, the database is referred to as DB.

First, the service history DB 151, the worker DB 152, the service target DB 153, and the incident DB 154 will be described.
The service history DB 151 stores service history data related to work that has been performed (that is, service execution results). For example, in the service history DB 151, information indicating a work start date / time, a work end date / time, an object, work content, and a worker is recorded for each work that has been performed. An example of the service history DB 151 is shown in FIG. As shown in FIG. 2, the service history DB 151 stores data relating to one work as one record, and includes a history ID, work start date / time, work end date / time, object ID, business classification, small category, and standard work per person. Eight fields of man-hour and worker ID are included. The history ID is an identifier of the record, that is, information for identifying the record. The object ID is an identifier of the object (in this embodiment, an elevator). The business classification indicates the type of work. The small category is a subdivision of business classification. The standard work man-hour per person represents the difficulty of work. In the example of FIG. 2, the standard work man-hour per person is represented by an estimated work time (in minutes) required when a standard worker works alone. The worker ID is an identifier of the worker.

  In the record whose history ID is h1000200, work start time: 2009/6/5 10:00: 00, work end time: 2009/6/5 10:45:00, object ID: b02, work classification: inspection, Small category: periodic inspection, standard work man-hour per person: 60 (minutes), worker ID: p1. There are two records whose history ID is h1000205, which indicates that two workers (workers p4 and p5) are jointly performing one work.

  The worker DB 152 stores worker data related to workers engaged in elevator maintenance services. For example, in the worker DB 152, information indicating a name, a title, and an organization (for example, a sales office) to which each worker is associated is recorded. An example of the worker DB 152 is shown in FIG. As shown in FIG. 3, the worker DB 152 stores data relating to one worker as one code, and includes five fields of worker ID, name, gender, job title, and sales office ID. The business office ID is an identifier of the business office to which the worker belongs.

  In the record whose worker ID is p1, the name is: maintenance Taro, gender: M, title: chief, and sales office ID: e10334. In the example of FIG. 3, the positions are classified into three levels, chief, general, and newcomer. The position can be used when calculating the amount of work to be described later, such as increasing the work time (also referred to as work man-hour) by a factor of 1.5 if it is the chief. Note that the sales office ID may not be included in the worker DB 152. Even if the sales office ID is not included in the worker DB 152, the sales office of the worker can be extracted using the service history DB 151, the worker DB 152, and the service target DB 153.

  The service target DB 153 stores service target data related to a target for providing a service. For example, in the service target DB 153, information on the type, position, and jurisdiction (or sales office) is recorded for each target. An example of the service target DB 153 is shown in FIG. As shown in FIG. 4, the service target DB 153 stores data related to one target as one record, and includes five fields of target ID, model, address, sales office ID, and contract type. The address indicates the position of the object. In this embodiment, the address is the address of the building where the elevator is installed.

  In the record whose object ID is b01, the model is k001, the address is 1-1 Nakahara-ku, Kawasaki city, the sales office ID is e10334, and the contract type is contract A. The object b01 (the object whose object ID is b01) is managed by the sales office e10334 (the sales office whose sales office ID is e10334). This does not mean that the object b01 is worked only by the workers belonging to the sales office e10334. The object b01 is mainly worked by a worker at the sales office e10334, but may be worked by a worker at another sales office. The model and contract type can be used when, for example, a work situation described later is analyzed for each model. The sales office ID may not be included in the service target DB 153. Even if the sales office ID is not included in the service target DB 153, the office to which the worker belongs can be extracted using the service history DB 151, the worker DB 152, and the service target DB 153.

  In the incident DB 154, incident data related to work that has been performed, which is work for dealing with an incident that has occurred in the object, is stored. Here, an incident refers to an event to be evaluated for service quality. In the example of the facility maintenance service, the incident includes, for example, an abnormal event (for example, failure) that occurs suddenly during operation of the facility. An example of the incident DB 154 is shown in FIG. As shown in FIG. 5, the incident DB 154 stores data relating to one incident as one record, and includes an object ID, an incident ID, a work start date and time, a work end date and time, a severity, a business classification, a small category, and a worker ID. 8 fields are included. The incident ID is an identifier of the incident. The work start date and time and the work end date and time represent the date and time when the work for dealing with the incident was started and ended, and the worker ID is an identifier of the worker who performed the work.

  In the record whose incident ID is inc20015, object ID: b01, work start date / time: 2012/7/18 10:00: 00, work end date / time: 2012/7/18 13:00:00, severity: S, Business classification: maintenance, small category: parts replacement, worker ID: p1. Severity can be used when priority is given to high-severity incidents when performing situation analysis. In the example of FIG. 5, the severity increases in the order of B, A, and S.

  The above-described DBs 151 to 154 are updated as appropriate. In one example, the DBs 151 to 154 are held in a server, and the DBs 151 to 154 are updated when an operator operates a terminal device (client) installed in each sales office. 1 illustrates an example in which the DBs 151 to 154 are provided outside the service analysis apparatus 100, the service analysis apparatus 100 may include the DBs 151 to 154.

Next, the configuration of the service analysis apparatus 100 shown in FIG. 1 will be described.
The work amount calculation unit 101 uses the service history DB 151 to calculate the amount of work performed on the target object for each target object. The amount of work is calculated based on, for example, the work time that is the difference between the work start date and time and the work end date and time stored in the service history DB 151. When the work is performed for each object a plurality of times, the work amount is calculated based on the total work time related to the work. The amount of work increases as the working time increases. The work amount calculation unit 101 may calculate the work time using a weighting factor according to the difficulty level of the work, the number of people who performed the work at the same time, the position of the worker, and the like.

  When using a weighting factor according to the job title, the work amount calculation unit 101 accesses the worker DB 152. If a worker whose title is the chief and a worker whose title is general work for the same amount of time, the amount of work is considered to be greater for the worker whose title is the chief. In one example, the general weighting factor is 1.0 and the chief weighting factor is 1.5. In this example, when a chief worker performs a work for 60 minutes on a certain object and a general worker performs a work for 50 minutes, the total work time used for calculating the work amount is 140 minutes (= 60 × 1.5 + 50 × 1.0).

An example of the work amount calculated for each object by the work amount calculation unit 101 is shown in FIG. In the example of FIG. 6, the workload is shown for nine objects b01 to b09. For example, the work amount of the object b01 is 300. In this embodiment, the work amount for each object is represented by a vector. In the example of FIG. 6, the work amount vector V can be expressed as follows.
V = (300, 310, 305, 200, 205, 215, 200, 210, 220)
In this example, the i-th element of the work amount vector V represents the work amount of the object b0i, where i is an integer from 1 to 9. For example, the first element of the work amount vector V represents the work amount of the object b01.

  The clustering unit 102 uses the service history DB 151, the worker DB 152, and the service target DB 153 to indicate affiliation information indicating the degree to which each worker belongs to each execution unit and the degree to which each execution unit manages each object. Generate management information. The clustering unit 102 groups a plurality of workers into a plurality of implementation units using a clustering method in order to generate affiliation information and management information. Specifically, the clustering unit 102 includes a work matrix calculation unit 103, an implementation unit clustering unit 104, and a normalization unit 105. Note that the clustering unit 102 can also generate affiliation information and management information using the service history DB 151 without referring to the worker DB 152 and the service target DB 153.

  The work matrix calculation unit 103 uses the service history DB 151, the worker DB 152, and the service target DB 153, and the amount of work performed by each worker on each object, that is, each worker is involved in each object. A work matrix that is a matrix representing the degree is calculated. An example of the work matrix is shown in FIG. FIG. 7 shows an example of a work matrix obtained when ten workers p1 to p10 work on nine objects b01 to b09. In the example of FIG. 7, the objects b01 to b03 are operated by the workers p1 to p3, the objects b04 to b06 are operated by the workers p4 to p6, and the objects b07 to b10 are operated by the workers p7 to p7. Work is being performed by p9. The amount of work stored in each element of the work matrix can be calculated based on the work time, the number of works, both the work time and the number of works. In the case where each worker performs work on each target object a plurality of times, the work time uses the total work time related to those work. The amount of work becomes larger as the working time is longer or the number of work is larger. The work matrix calculation unit 103 may calculate the work amount by performing weighting according to the difficulty level of the work, the number of people who performed the work at the same time, the position of the worker, and the like.

  The execution unit clustering unit 104 clusters the work matrix calculated by the work matrix calculation unit 103, thereby generating a first matrix indicating the relationship between the object and the execution unit and a second matrix indicating the relationship between the worker and the execution unit. Generate at least two matrices containing. Workers are grouped into multiple implementation units by clustering. As a clustering method, for example, a method called co-clustering can be used.

FIG. 8 shows the result of co-clustering on the work matrix of FIG. In the example of FIG. 8, the number of clusters is 3, and clustering is performed according to co-clustering using non-negative three-factor matrix decomposition. The number of clusters corresponds to the number of implementing units. In this example, the workers are grouped into three execution units u1, u2, u3. Working matrix X has three matrix F by co-clustering, S, is decomposed into ^{G T.} G ^T represents a transposed matrix of the matrix G. The matrix F corresponds to the result of clustering the relationship between the object and the execution unit, that is, the first matrix. Matrix G ^T is, as a result of the clustering relationship between execution units and the operator, i.e., corresponding to the second matrix.

  In the matrix F, the degree of management of the nine objects b01 to b09 by the three execution units u1, u2, and u3 is shown. In the matrix F, the first to ninth rows correspond to the workers b1 to b09, respectively. Here, it is assumed that the first, second, and third rows of the matrix F correspond to the execution units u3, u2, and u1, respectively. In the first to third rows, only the first column contains a value greater than zero. This represents that the objects b01 to b03 are managed only by the execution unit u3. Similarly, the objects b04 to b06 are managed only by the execution unit u2, and the objects b07 to b09 are managed only by the execution unit u1.

The matrix ^{G T,} the 10 worker P1 to P10, the degree of affiliation is shown to three execution units u1, u2, u3. In the matrix ^{G T,} the third row from the first row correspond to the execution units u1, u2, u3, 10th column from the first column correspond to the worker p10 from the worker p1. In the first to third columns, only the third row has a value greater than zero. This indicates that the workers p1 to p3 belong only to the execution unit u3. Similarly, the workers p4 to b6 belong only to the execution unit u2, and the workers p7 to p10 belong to only the execution unit u1.

  The above example is a simple decomposition for small-scale data in which both the number of objects and the number of workers are small. However, when using co-clustering, large-scale data can be decomposed. Although the presence or absence of management or affiliation can be determined by these matrices F and G, the exact degree of management or affiliation is not known. Post-processing of the clustering result is necessary to estimate the degree.

  The normalization unit 105 normalizes and manages the two clustering results obtained by the implementation unit clustering unit 104, that is, the clustering result about the relationship between the object and the implementation unit and the clustering result about the relationship between the worker and the implementation unit. Two matrices, a matrix and an affiliation matrix, are calculated. The affiliation matrix is a matrix that represents the proportion of workers p1 to p10 belonging to the execution units u1, u2, and u3. An example of the affiliation matrix is shown in FIG. The affiliation matrix is normalized to a probability value such that the sum of the values of the execution units u1 to u3 for each worker is 1.0. In the example of FIG. 9, the sum of the three elements in each row is 1.0. The affiliation matrix is stored in the affiliation information storage unit 107 as affiliation information.

  The management matrix is a matrix that represents the rate at which the objects b01 to b09 are managed by the execution units u1 to u3. An example of the management matrix is shown in FIG. The management matrix is normalized to a probability value such that the sum of the values of the execution units u1 to u3 is 1.0 for each object. In the example of FIG. 10, the sum of the three elements in each row is 1.0. In the example of FIG. 10, for example, the object b05 is managed only by the execution unit u2. The management matrix is stored in the management information storage unit 108 as management information.

The incident amount calculation unit 106 calculates the incident amount of each target using the incident DB 154. Here, as the incident amount, downtime, incident response time, incident occurrence rate, and the like are used. The incident response time is the difference between the work start date and time and the work end date and time. When an incident has occurred multiple times in each object, the incident response time uses the total of the work time related to the work for responding to those incidents. When downtime is used as the incident amount, the incident data includes a downtime field. An example of the calculation result of the incident amount calculation unit 106 is shown in FIG. FIG. 11 shows incident amounts for nine objects b01 to b09. In the present embodiment, the incident amount of each object is represented by a vector. In the example of FIG. 11, the incident amount vector W can be expressed as follows.
W = (2.2, 2.0, 2.1, 1.7, 1.8, 1.8, 1.4, 1.6, 1.5)
In this example, the i-th element of the incident amount vector W represents the work amount of the object b0i, where i is an integer from 1 to 9. For example, the first element of the work amount vector V represents the work amount of the object b01.

  The situation analysis unit 109 analyzes the incident situation and the work situation using the work amount for each target object, the incident amount for each target object, affiliation information, and management information. In the present embodiment, the work amount for each object, the incident amount for each object, affiliation information, and management information correspond to the work amount vector, the incident amount vector, the affiliation matrix, and the management matrix. Specifically, the situation analysis unit 109 includes a work situation analysis unit 110, an incident situation analysis unit 111, and a relationship analysis unit 112.

  The work situation analysis unit 110 performs an analysis on the work situation for each execution unit based on the work amount vector, the affiliation matrix, and the management matrix. Examples of the work situation include a work load per worker and a work amount per object.

  The incident status analysis unit 111 analyzes the incident status for each execution unit based on the incident amount vector and the management matrix. Examples of the incident status include, for example, an incident occurrence amount, an incident occurrence rate, and the like.

  Based on the analysis result obtained by the work situation analysis unit 110 and the analysis result obtained by the incident situation analysis unit 111, the relationship analysis unit 112 determines the relationship such as the correlation between the workload and the incident and the correlation between the work amount and the incident. analyse.

  An example of analysis results obtained by the work situation analysis unit 110 and the incident situation analysis unit 111 is shown in FIG. In FIG. 12, X1 indicates the work load per worker, X2 indicates the work amount per object, and Y indicates the incident occurrence rate. In the example of FIG. 12, the workload X1 per worker is 158 for the execution unit u1, 207 for the execution unit u2, and 305 for the execution unit u3. The work amount X2 per object is 210 for the execution unit u1, 207 for the execution unit u2, and 305 for the execution unit u3. Further, the incident occurrence rate Y is 1.5 for the execution unit u1, 1.8 for the execution unit u2, and 2.1 for the execution unit u3.

  In this embodiment, the calculation method of X1, X2, and Y follows the following mathematical formula.

X1 = {P ^T × W} _i / {Q ^T × 1} _i (1)
X2 = {P ^T × W} _i / {P ^T × 1} _i (2)
Y = {P ^T × V} _i / {P ^T × 1} _i (3)
Here, W represents a work amount vector, V represents an incident amount vector, P represents a management matrix, and Q represents an affiliation matrix. A ^T represents a transposed matrix of the matrix A. “1” in Equation (1) is a vector having the same number of elements as the number of columns of the transposed matrix of the affiliation matrix Q, and all the elements are 1. “1” in the mathematical expressions (2) and (3) is a vector having the same number of elements as the number of columns of the transposed matrix of the management matrix P, and all the elements are 1. {B} _i / {C} _i represents an operation of dividing the i-th element of the vector B by the i-th element of the vector C.

  In calculating the work load X1 per worker, first, two vectors are calculated. The first is a vector calculated by the matrix product of the transposed matrix of the management matrix P and the work amount vector W, and the second is calculated by the matrix product of the transposed matrix of the belonging matrix Q and the vector 1. Is a vector. Next, X1 is obtained by dividing each element of the former vector by each element of the latter vector.

  In calculating the work amount X2 per object, first, two vectors are calculated. The first is a vector calculated by the matrix product of the transposed matrix of the management matrix P and the work amount vector W, and the second is calculated by the matrix product of the transposed matrix of the management matrix P and the vector 1. Is a vector. Next, X2 is obtained by dividing each element of the former vector by each element of the latter vector.

  In calculating the incident occurrence rate Y, first, two vectors are calculated. The first is a vector calculated by the matrix product of the transposed matrix of the management matrix P and the incident amount vector V, and the second is calculated by the matrix product of the transposed matrix of the management matrix P and the vector 1. Is a vector. Next, Y is obtained by dividing each element of the former vector by each element of the latter vector.

  FIGS. 13A and 13B show examples in which a relationship analysis is performed using the analysis results shown in FIG. FIG. 13A shows the result of analyzing the correlation between the workload X1 per worker and the incident rate Y. In FIG. 13A, the horizontal axis represents the workload X1 per worker, and the vertical axis represents the incident occurrence rate Y. FIG. 13B shows the result of analyzing the correlation between the work amount X2 per object and the incident occurrence rate Y. In FIG. 13B, the horizontal axis represents the work amount X2 per object, and the vertical axis represents the incident occurrence rate Y.

  From FIG. 13 (a), it can be seen that there is a correlation that the incident occurrence rate increases as the workload per worker increases. From FIG. 13 (b), it can be seen that the execution unit u3 has a relatively high work amount per object and a high incident occurrence rate, so that there is not enough manpower for the large work amount. Thereby, an analyst (for example, an administrator of an elevator maintenance service) can come up with a countermeasure to increase the number of personnel in order to solve this problem. Furthermore, from FIG. 13B, it can be seen that when the execution units u1 and u2 are compared, the amount of work is the same, but the incident occurrence rate of the execution unit u2 is higher. Thereby, the analyst can recognize that there is a possibility of reducing the incident occurrence rate of the execution unit u2.

  The analysis result output unit 113 outputs the analysis result obtained by the situation analysis unit 109. The analysis result is presented to the user in the form of a graph or a table, for example. The presentation method may be executed by any method such as image output by a display device (not shown) or print output by a printer device (not shown). For example, the relationship between the workload per worker and the incident occurrence rate as in the graph shown in FIG. 13A, and the relationship between the work amount per object and the incident occurrence rate shown in FIG. The data necessary for these analyzes (for example, data constituting the table shown in FIG. 12) is output.

  The input unit 114 receives input of analysis conditions from an analyst as an analysis query. For example, the analyst can use the input unit 114 to specify the time range of data used for analysis, the number of clusters, and the like.

  The service analysis apparatus 100 including the elements 101 to 114 described above can evaluate service quality for each implementation unit by grouping a plurality of workers into a plurality of implementation units. Grouping can be performed by clustering on the matrix. In this case, analysis can be performed at high speed. Furthermore, large-scale data can be handled by using co-clustering as a clustering method.

Next, the operation of the service analysis apparatus 100 shown in FIG. 1 will be described.
FIG. 14 shows an example of the analysis processing procedure of the service analysis apparatus 100. First, the user uses the input unit 114 to specify an analysis target period, the number of clusters, and the like as an analysis query. Thereby, the analysis process starts. In step S1401 of FIG. 14, the service analysis apparatus 100 reads the service history DB 151, the worker DB 152, the service target DB 153, the incident DB 154, and the analysis parameters.

  In step S1402, the work amount calculation unit 101 uses the service history DB 151 to calculate the work amount for each object and generate a work amount vector. In step S1403, the incident amount calculation unit 106 uses the incident DB 154 to calculate an incident amount for each target and generate an incident amount vector.

  In step S <b> 1404, the work matrix calculation unit 103 generates a work matrix using the service history DB 151, the worker DB 152, and the service target DB 153. In step S1405, the execution unit clustering unit 104 performs clustering on the target object and the execution unit. In step S1406, the normalization unit 105 performs normalization processing on the matrix obtained as a result of clustering, and generates an affiliation matrix and a management matrix.

  In step S1407, the work situation analysis unit 110 performs an analysis on the work situation based on the work amount vector, the affiliation matrix, and the management matrix, and the incident situation analysis unit 111 performs an incident based on the incident amount vector and the management matrix. Analyze the situation. In step S1408, the relationship analysis unit 112 analyzes the relationship between the work status and the incident status. In step S1408, the analysis result output unit 113 outputs the analysis result of the relationship analysis unit 112.

  The analysis processing procedure shown in FIG. 14 is an example, and the processing steps may be executed in a different order from the analysis processing procedure shown in FIG. For example, the process of step S1402 and the process of step S1403 may be executed after the process of step S1406, or may be executed in parallel with a series of processes shown in steps S1404 to S1406.

  Next, the clustering process shown in step S1405 of FIG. 14 will be described in detail. The processing in step S1404 is part of the processing performed by the execution unit clustering unit 104. FIG. 15 shows a processing procedure when a co-clustering method called non-negative three-factor matrix decomposition is used. Other clustering algorithms such as the k-means method can also be used. However, by using the co-clustering method, it becomes possible to deal with large-scale data.

In step S1501 of FIG. 15, the implementation unit clustering unit 104 reads the matrix X to be clustered. Here, the matrix X is a work matrix generated by the work matrix calculation unit 103. In step S1502, the implementation unit clustering unit 104 sets initial values of G _jk , F _ik , and S _ik using random values, and further assigns zero to the count variable N.

In steps S1503 to S1505, the implementation unit clustering unit 104 calculates a value that is a result of the non-negative three-factor matrix decomposition according to each calculation formula. Specifically, the implementation unit clustering unit 104 substitutes a value calculated by the following equation (4) for G _jk (step S1503), and substitutes a value calculated by the following equation (5) for S _ik ( step _S1504), and substitutes the value calculated by the following equation (6) to _{S ik} (step S1505).

In step S1506, execution units clustering unit 104, according to the following equation (7) to calculate the distance dist is the square of the difference between the working matrix X with a matrix product FSG ^T.

In step S1507, the count variable N is incremented, that is, incremented by one. In step S1508, it is determined whether at least one of the condition that the distance dist is smaller than a predetermined value (for example, 10) and the condition that the count variable N is larger than a predetermined value (for example, 1200) is satisfied. If both conditions are not satisfied, the process returns to step S1503; otherwise, the process proceeds to step S1509. In step S1510, the implementation unit clustering unit 104 outputs the calculation results of the matrices G, F, and S.

  In the example of FIG. 15, the predetermined value for the count variable N is 1200, and the predetermined value for the distance dist is 10, but any value can be set for these predetermined values. The numerical example of the present embodiment shows an example in which matrix decomposition is performed using this parameter.

Matrix F, G ^T obtained in the process shown in FIG. 15 is not a probability value. In the present embodiment, to convert these matrices F, the G ^T in the probability. As shown in FIG. 16, the matrix X is 3 factors F, S, is approximated by the product of ^{G T.} If the matrix X is working matrix, factor F corresponds to a matrix representing the relation between execution units and objects, factor G ^T corresponding to the matrix representing the relation between the execution units and the operator. Referring to FIGS. 16 and 17, calculates a management matrix P from the matrix F and S, explaining the process of calculating the belonging matrix Q from G ^T.

  In step S1701 of FIG. 17, the execution unit clustering unit 104 reads the work matrix X. In step S1702, the execution unit clustering unit 104 performs co-clustering on the work matrix X to obtain three factors F, S, and G as shown in FIG. In the service analysis, the matrix F is information indicating a spatial area to which each object belongs, and can be regarded as information indicating whether the implementation unit is in charge of the matrix S for the area. Further, the matrix G can be regarded as information indicating the presence / absence of belonging to the execution unit.

  In step S1703 of FIG. 17, the implementation unit clustering unit 104 corrects the norm of each column of the matrix G to obtain a matrix G ′. Subsequently, the implementation unit clustering unit 104 normalizes the rows of the matrix G ′ (step S1704), and sets the matrix obtained thereby as the belonging matrix Q (step S1705). In the normalization processing of the matrix after decomposition as shown in FIG. 16, normalization is performed using the row elements after performing norm correction.

  In step S1706, the implementation unit clustering unit 104 binarizes each element of the matrix F with {0, 1} to obtain a matrix F ′. In step S1707, the implementation unit clustering unit 104 obtains a matrix S ′ that satisfies FS≈F ′S ′. Subsequently, the implementation unit clustering unit 104 normalizes the rows of the matrix product F ′S ′ (step S1708), and sets the matrix obtained thereby as the management matrix P (step S1709).

  In this way, the membership matrix and the management matrix can be generated from the work matrix. Note that the procedure shown in FIG. 17 is an example, and the processing steps may be executed in a different order from the procedure shown in FIG. For example, the series of processes shown in steps S1703 to S1705 may be executed after the series of processes shown in steps S1706 to S1709, or may be executed in parallel with the series of processes shown in steps S1706 to S1709. Also good.

  In the above-described example, the object is completely divided and managed for each execution unit, and the worker belongs to each execution unit. That is, each object is managed by one execution unit, and each worker belongs to one execution unit. Each object may be managed by a plurality of execution units, and each worker may belong to a plurality of execution units.

  FIG. 18 shows another example of a work matrix, and a management matrix and an affiliation matrix calculated from the work matrix. The management matrix and affiliation matrix in FIG. 18 are obtained by clustering the work matrix in FIG. Referring to the work matrix in FIG. 18, the objects b01 to b03 are worked by the workers p1 to p3, the objects b04 to b06 are worked by the workers p4 to p10, and the objects b07 to b09 are the workers p7 to p10. Is being worked on by. Referring to the management matrix, the objects b01 to b03 are managed only by the execution unit u3, the objects b04 to b06 are managed by the two execution units u2 and u3, and the objects b07 to b09 are managed only by the execution unit u2. ing. For example, the object b04 is managed by the execution units u2 and u3 at a ratio of 0.734 to 0.266. Further, referring to the affiliation matrix, the workers p1 to p3 belong only to the execution unit u3, and the workers p4 to p10 belong to the two execution units u1 and u2. For example, the worker p4 belongs to the execution units u1 and u2 at a ratio of 0.999 to 0.001.

  In this way, in a service where a plurality of workers work on a plurality of objects, even when the worker performs support work across organizations (sales offices), the execution unit is extracted from the service history data. Information for analyzing the relationship between the object and the execution unit and the relationship between the worker and the execution unit can be obtained.

  As described above, the service analysis apparatus according to the present embodiment can evaluate service quality for each implementation unit by grouping a plurality of workers into a plurality of implementation units by clustering. Furthermore, by using matrix calculation, it is possible to perform analysis at high speed even for large-scale data. In the example of equipment maintenance service, it is possible to accurately grasp the failure occurrence rate, workload, work skill, etc. for each execution unit from the analysis results, reducing failure, smoothing the workload, improving the efficiency of human resource development, etc. Can be realized.

  In the above-described embodiment, an example of an example of an elevator maintenance service has been described. However, the present invention is not limited to this, and the service analysis apparatus is applicable to any service in which a plurality of persons work on a plurality of objects. can do. Examples of such services include nursing services such as day services and medical nursing. In the example of the care service, the relationship between the user and the worker can be analyzed. In the case of medical nursing, it is possible to analyze the relationship between a patient and a nurse.

  Furthermore, since the analysis target period can be selected using the input unit 114, time series analysis can be performed by appropriately setting the analysis target period. For example, if the workload per person in the implementation unit is calculated every year and the variation is calculated in time series, it can be confirmed whether a mismatch between the workload and the worker has occurred.

  The instructions shown in the processing procedure shown in the above-described embodiment can be executed based on a program that is software. The general-purpose computer system stores this program in advance and reads this program, so that the same effect as that obtained by the service analysis apparatus 100 of the above-described embodiment can be obtained.

  FIG. 19 schematically shows a hardware configuration example of the service analysis apparatus 100 shown in FIG. The service analysis apparatus 100 includes a CPU (Central Processing Unit) 1901, a RAM (Random Access Memory) 1902, an HDD (Hard Disk Drive) 1903, a graphic processing apparatus 1904, an input interface 1905, and a communication interface 1906.

  CPU 1901 operates according to a program stored in HDD 1903 or ROM (not shown). The RAM 1902 stores data necessary for the CPU 1901 to execute various processes as necessary. The HDD 1903 stores programs executed by the CPU 1901, data necessary for the CPU 1901 to execute various processes, and the like. The graphic processing device 1904 displays an image corresponding to the image data given from the CPU 1901 on the monitor 1908. A keyboard 1909 and a mouse 1910 are connected to the input interface 1905. The input interface 1905 gives a signal to the CPU 1901 according to the operation of the keyboard 1909 and the mouse 1910 by the analyst (user). The communication interface 1906 accesses a database (for example, the DBs 151 to 154 shown in FIG. 1) via the LAN 1907, and data obtained from the database is temporarily stored in the RAM 1902.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 100 ... Service analysis apparatus, 101 ... Work amount calculation part, 102 ... Clustering part, 103 ... Work matrix calculation part, 104 ... Implementation unit clustering part, 105 ... Normalization part, 106 ... Incident amount calculation part, 107 ... Affiliation information storage 108 ... Management matrix storage unit 109 ... Situation analysis unit 110 ... Work situation analysis unit 111 ... Incident situation analysis unit 112 ... Relationship analysis unit 113 ... Analysis result output unit 1901 ... CPU 1902 ... RAM, DESCRIPTION OF SYMBOLS 1903 ... HDD, 1904 ... Graphic processing device, 1905 ... Input interface, 1906 ... Communication interface, 1907 ... LAN, 1908 ... Monitor, 1909 ... Keyboard, 1910 ... Mouse.

Claims (8)

A service analyzer for evaluating the quality of a service in which a plurality of workers perform work on a plurality of objects,
A work amount calculation unit that calculates a work amount for each of the plurality of objects based on service history data relating to work that has been performed;
Based on the service history data, management information indicating the degree to which each of the plurality of execution units that group the plurality of workers manages the plurality of objects, and each of the plurality of workers includes the plurality of A clustering unit for generating affiliation information indicating the degree of belonging to each of the implementation units;
An incident amount calculation unit for calculating an incident amount for each of the plurality of objects, based on incident data related to the performed work, for dealing with incidents occurring in the plurality of objects;
Based on the work amount for each of the plurality of objects, the affiliation information, the management information, and the incident amount for each of the plurality of objects, the work status for each of the plurality of execution units and the incident for each of the plurality of execution units A situation analysis section that analyzes the situation;
Service analysis apparatus comprising: The clustering unit
Based on the service history database, a work matrix calculating unit that calculates a work matrix indicating a degree to which each of the plurality of workers is involved in each of the plurality of service objects;
By performing clustering on the work matrix, the first clustering result indicating the relationship between the plurality of workers and the plurality of execution units and the relationship between the plurality of objects and the plurality of execution units are shown. An implementation unit clustering unit for generating a second clustering result;
A normalization unit that generates the affiliation information by performing a normalization process on the first clustering result, and generates the management information by performing a normalization process on the second clustering result;
The service analysis device according to claim 1, comprising: The implementation unit clustering unit decomposes the work matrix into three matrices by co-clustering processing,
The service analysis apparatus according to claim 2, wherein the normalization unit generates the affiliation information and the management information from the three matrices. The situation analysis unit
From a work amount for each of the plurality of objects, the affiliation information, and the management information, a work situation analysis unit that analyzes a work situation for each of the plurality of execution units;
From the management information and the incident amount for each of the plurality of objects, an incident status analysis unit that analyzes the incident status for each of the plurality of implementation units;
The service analysis apparatus according to any one of claims 1 to 3, further comprising:   The clustering unit generates the affiliation information and the management information based on the service history data, worker data regarding the plurality of workers, and target data regarding the plurality of targets. Service analysis equipment.   The service analysis apparatus according to claim 1, wherein the service is a facility maintenance service.   The service analysis apparatus according to claim 6, wherein the plurality of objects are the plurality of facilities to be maintained. A service analysis method for evaluating the quality of a service in which a plurality of workers perform work on a plurality of objects,
Calculating an amount of work for each of the plurality of objects based on service history data relating to work that has been performed;
Based on the service history data, affiliation information indicating the degree to which each of the plurality of workers belongs to each of a plurality of execution units in which the plurality of workers are grouped, and each of the plurality of execution units Generating management information indicating a degree of managing each of the plurality of objects;
Calculating the amount of incident for each of the plurality of objects based on incident data related to the work that has been performed to deal with incidents occurring in the plurality of objects;
Based on the work amount for each of the plurality of objects, the affiliation information, the management information, and the incident amount for each of the plurality of objects, the work status for each of the plurality of execution units and the incident for each of the plurality of execution units Analyzing the situation,
A service analysis method comprising:

Images