Patent Application
20230419219
This application is based upon and claims the benefit of priority of the prior Japanese Patent application No. 2022-103368, filed on Jun. 28, 2022, the entire contents of which are incorporated herein by reference.
The embodiment discussed herein relates to a method for generating a workflow and a computer-readable recording medium having stored therein a workflow generating program.
In the practice of policy planning, policy plans are prepared based on consideration, experience, and assumptions, which makes it difficult to achieve the policy index (Key Performance Indicator: KPI) target. For the above, demands have been arisen for generating an achievable policy based on evidence-based KPI prediction at a policy planning stage and presenting the generated policy to a policy developer in advance.
Conventionally, in regard of a policy candidate, for example, a technique has been proposed which predicts a KPI of a policy from integrated data including various types of data and supports decision-making in policy planning.
For example, as integrated data, a history of daily health-information, a history of diagnosis results, a history of lifestyle habits, and a history of examination results for each individual local resident are prepared in advance.
Then, as policy KPI prediction, “when a policy candidate is executed, the probability of becoming obese within a predetermined period is predicted (at three levels of high, medium, and low)” for individuals in the target regional area.
In setting a policy candidate, for example, a policy goal of “reducing the prevalence of obesity in adults in the target regional area to 25%”, a policy candidate of “implementing exercise and nutritional guidance by physicians for young people living in the regional area”, and a KPI of “prevalence of obesity in adults in the regional area” are set.
However, in such a conventional method for aiding decision-making on policy planning, it is difficult to generate a new policy when a KPI of a policy is predicted and it is determined that the policy is unable to reach the goal. For example, when the predicted result of KPI “prevalence of obesity in adults” is 30% in the original policy candidate, the policy developer needs to reconsider a policy candidate in person because the predicted result does not reach KPI target of 25%, which makes the development complicated.
Therefore, a method of regenerating a new policy in case where the original policy does not reach the goal is also known. For example, a synthesized policy is generated by combining multiple policy candidates, and such synthesized policies are listed as many as conceivable. For example, if N policy candidates are present, N-th power of two synthesized policies are generated. Then, KPI prediction is sequentially performed on each of the synthesized policies, and when a policy candidate that can reach the goal is found, the policy candidate is adopted.
However, in this method, an increase in policy candidates exponentially increases combinations of synthesized policies. Therefore, since KPI prediction of all synthesized policy is simulated, the computational load is increased.
In particular, when many parameters that can be varied in the policy are present, combination of synthesized policy becomes enormous and the computational load for simulating KPI prediction increases to an unrealistic level.
According to an aspect of the embodiments, a computer-implemented method for generating a workflow includes: generating a plurality of candidate workflows by making a modification to part of an existing workflow, the existing workflow defining a plurality of condition branches and following contents at respective destinations of the plurality of condition branches; obtaining a plurality of KPI (Key Performance Indicator) fluctuation vectors one for each of the plurality of candidate workflows from the existing workflow; generating a plurality of synthesized policies by synthesizing two or more of the plurality of KPI fluctuation vectors; calculating, when one of selected routes of the plurality of candidate workflows contains a plurality of the modifications, a KPI predicted value of each of the plurality of synthesized policies by reflecting a weight parameter on the KPI fluctuation vector, the weight parameter being set such that a weight for a modification to an upper point of the selected route among the plurality of modifications is higher than a weight for a modification to a lower point of the selected route among the plurality of modifications; and outputting a workflow of one of the plurality of synthesized policies the KPI predicted value of which satisfies a target value.
The object and advantages of the invention will be realized and attained by means of the elements and combinations particularly pointed out in the claims.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention, as claimed.
The following method can be conceived as a method of performing KPI prediction of synthesized policy described above.
First, KPI prediction is performed on an existing policy (one piece) and policy candidates (N pieces at the maximum) serving as sources of synthesis. In
Next, a vector (referred to as a KPI fluctuation vector) indicating a difference between a KPI predicted value of the existing policy and a KPI predicted value of each policy candidate is obtained.
In
A value obtained by simply adding respective KPI fluctuation vectors of the policy candidates to the KPI predicted value of the existing policy is regarded as a predicted value of the synthesized policy.
In
Here, the policy may be represented in the form of a workflow in which multiple condition distributions and multiple nodes are combined. In addition, a workflow representing a policy may be referred to as a policy workflow. A policy workflow may include an element except for a condition branch and a node.
In
In the specific medical checkup (specified checkup), for example, the respective determination branches for results of tests such as “whether the eGFR test value is less than a predetermined threshold (for example, 50 ml/min/1.73 m2)” and “whether the urinary protein is 2+ or more” correspond to condition branches in the workflow.
In addition, the “treatment by a nephrologist”, “health guidance by a primary care physician”, and “treatment by a diabetic specialist” specified as the results of the respective determination branches correspond to the nodes in the workflow. In these nodes, the nodes may be referred to as intervening nodes because they are accompanied by intervention by any of a nephrologist, a primary care physician, and a diabetic specialist.
In the policy workflow, multiple condition branch are sequentially traced based on individual state data of the individual of a target to the policy, and nodes allocated to the individual are determined. A node corresponds to a following content reached at a branch destination in a workflow.
In
The policy candidate 1 indicated by the reference symbol B is obtained by adding a perturbation A to the existing policy, and the policy candidate 2 indicated by the reference symbol C is obtained by adding a perturbation B to existing policy. The synthesized policy indicated by the reference symbol D is a combination of the policy candidate 1 and the policy candidate 2.
A perturbation is a predetermined change given to the policy workflow within a predetermined range. A perturbation may be given to the element (e.g., a condition branch, or a node) that constitute a policy workflow. The predetermined change may be, for example, an increase in the criterion parameters of a condition.
As a premise, in order to perform KPI prediction using a policy workflow, a workflow is applied to all the policy target persons, the number of persons to be allocated to each node is calculated, and KPI prediction is performed on the basis of results of allocation to the respective nodes. For example, KPI of the prevalence of obesity morbidity is increased if the number of persons allocated to the obesity guidance is small.
On the basis of the premise described above, as illustrated in
In the reference symbol B of
In the embodiment illustrated in
Therefore, it is reasonable to determine a KPI predicted value of the synthesized policy by adding a KPI predicted value of the existing policy and KPI fluctuation vectors of the policy candidates 1 and 2.
However, if the influence ranges of the respective perturbation overlap, the influence of an upper perturbation largely affects but the influence of a lower perturbation less affects, so that this method does not correctly function.
In the example illustrated in
The upstream side of the workflow is called upper, and the downstream side is called lower. In the embodiment illustrated in
Since, when the policy candidate 1 and the policy candidate 2 are synthesized, the policy candidate 1 on the upper point is dominant, the result of node allocation of the synthesized policy is not a simple sum of the respective difference of the results of node allocation of the respective policy candidates 1 and 2.
Therefore, the KPI predicted value of the synthesized policy is not a simple sum, and KPI prediction may be erroneously estimated even if the KPI predicted value of the existing policy and KPI fluctuation vectors of the policy candidate 1 and 2 are added.
Therefore, in an information processing apparatus 1 serving as an example of the present embodiment, when policy candidates to be synthesized are synthesized and the influence ranges of perturbations overlap, a more precise KPI can be predicted by considering the influence on a lower perturbation (at a lower point) from an upper perturbation (at an upper point).
Hereinafter, one embodiment of the present method and program for generating a workflow will now be described with reference to the accompanying drawings. However, the following embodiment is merely illustrative and is not intended to exclude the application of various modifications and techniques not explicitly described in the embodiment. Namely, the present embodiment can be variously modified and implemented without departing from the scope thereof. Further, each of the drawings can include additional functions not illustrated therein to the elements illustrated in the drawing.
The workflow generating apparatus 1 generates a workflow that can achieve a higher KPI on the basis of a workflow of an existing policy (existing policy workflow).
(A) Example of Hardware Configuration
The workflow generating apparatus 1 according to the one embodiment may be a virtual server (VM; Virtual Machine) or a physical server. The function of the workflow generating apparatus 1 may be achieved by one computer or by two or more computers. Further, at least some of the functions of the workflow generating apparatus 1 may be implemented using Hardware (HW) resources and Network (NW) resources provided by cloud environment.
As illustrated in
The processor 10a is an example of an arithmetic operation processing device that performs various controls and calculations. The processor 10a may be communicably connected to the blocks in the computer 10 via a bus 10j. The processor 10a may be a multiprocessor including multiple processors, may be a multicore processor having multiple processor cores, or may have a configuration having multiple multicore processors.
The processor 10a may be any one of integrated circuits (ICs) such as Central Processing Units (CPUs), Micro Processing Units (MPUs), Accelerated Processing Units (APUs), Digital Signal Processors (DSPs), Application Specific ICs (ASICs) and Field Programmable Gate Arrays (FPGAs), or combinations of two or more of these ICs.
The graphic processing device 10b executes a screen displaying control on an outputting device such as a monitor included in IO device 10f. Example of the graphic processing device 10b are various type of arithmetic operation processing apparatus, and include ICs such as Graphics Processing Units (GPUs, APUs, DSPs, ASICs, and FPGAs.
The memory 10c is an example of a HW device that stores information such as various types of data and programs. Examples of the memory 10c include one or both of a volatile memory such as a Dynamic Random Access Memory (DRAM) and a non-volatile memory such as a Persistent Memory (PM).
The storing device 10d is an example of a HW device that stores information such as various types of data and programs. Examples of the storing device 10d include a magnetic disk device such as a Hard Disk Drive (HDD), a semiconductor drive device such as a Solid State Drive (SSD), and various storing devices such as a non-volatile memory. Examples of the non-volatile memory include a flash memory, a Storage Class Memory (SCM), and a Read Only Memory (ROM).
The storing device 10d may store a program 10h (workflow generating program) that implements all or part of various functions of the computer 10.
For example, the processor 10a of the workflow generating apparatus 1 can achieve the functions of generating a workflow to be detailed below by expanding the program 10h stored in the storing device 10d onto the memory 10c and executing the expanded program 10h. In the storing device 10d, various data pieces generated in the course of processing performed by each element (see
The I/F device 10e is an example of a communication IF that controls connection and communication between the present computer 10 and another computer. For example, the I/F device 10e may include an applying adapter conforming to Local Area Network (LAN) such as Ethernet (registered trademark) or optical communication such as Fibre Channel (FC). The applying adapter may be compatible with one of or both wireless and wired communication schemes.
For example, the workflow generating apparatus 1 may be communicably connected, through the IF device 10e and a network, to another non-illustrated information processing apparatus. Furthermore, the program 10h may be downloaded from the network to the computer 10 through the communication IF and be stored in the storing device 10d, for example.
The IO device 10f may include one or both of an input device and an output device. Examples of the input device include a keyboard, a mouse, and a touch panel. Examples of the output device include a monitor, a projector, and a printer. The IO device 10f may include, for example, a touch panel that integrates an input device and an output device. The output device may be connected to the graphic processing device 10b.
The reader 10g is an example of a reader that reads data and programs recorded on a recording medium 10i. The reader 10g may include a connecting terminal or device to which the recording medium 10i can be connected or inserted. Examples of the reader 10g include an applying adapter conforming to, for example, Universal Serial Bus (USB), a drive apparatus that accesses a recording disk, and a card reader that accesses a flash memory such as an SD card. The program 10h may be stored in the recording medium 10i. The reader 10g may read the program 10h from the recording medium 10i and store the read program 10h into the storing device 10d.
The recording medium 10i is an example of a non-transitory computer-readable recording medium such as a magnetic/optical disk, and a flash memory. Examples of the magnetic/optical disk include a flexible disk, a Compact Disc (CD), a Digital Versatile Disc (DVD), a Blu-ray disk, and a Holographic Versatile Disc (HVD). Examples of the flash memory include a semiconductor memory such as a USB memory and an SD card.
The HW configuration of the computer 10 described above is exemplary. Accordingly, the computer 10 may appropriately undergo increase or decrease of HW devices (e.g., addition or deletion of arbitrary blocks), division, integration in an arbitrary combination, and addition or deletion of the bus.
(B) Example of Functional Configuration:
As illustrated in
The existing policy obtaining unit 101 obtains an existing policy. The existing policy obtaining unit 101 may obtain an existing policy that has been generated in any known method and stored in a predetermined storing region of the storing device 10d in advance, for example, by reading the existing policy. Alternatively, the existing policy obtaining unit 101 may obtain an existing policy from another computer connected via the IF device 10e.
A policy is expressed in the form of a workflow having multiple condition branches and multiple nodes (intervening nodes). A workflow representing a policy may be referred to as a policy workflow or a policy flow. An existing policy is also represented by a policy workflow.
For each individual target of a policy, intervening nodes allocated to the individual are determined by sequentially tracing the policy workflow through multiple condition branches on the basis of the individual state data.
In
The policy workflow has a tree structure, and the starting node side not having a parent (upper side in the drawing) may be referred to as the upstream side or upper side, and the terminal node side not having a child (lower side in the drawing) may be referred to as the downstream side or lower side.
In
The flow definition table has description of a branch destination and description of a branch condition. In the flow definition table indicated by the reference symbol B in
In the flow definition table indicated by reference symbol B in
In the present workflow generating apparatus 1, the information of the policy workflow may be managed by using such a flow definition table.
The existing policy obtaining unit 101 may obtain the flow definition table of an existing policy workflow together with the existing policy workflow. The existing policy obtaining unit 101 may generate a flow definition table based on the existing policy workflow.
The policy candidate generating unit 102 generates a policy candidate by adding a perturbation to an existing policy workflow.
The policy candidate generating unit 102 generates multiple policy candidates by providing such a perturbation to a condition branch or a node in an existing policy workflow. The method of providing a perturbation to each condition branch or each node is defined in advance.
The policy candidate generating unit 102 creates a policy candidate by, for example, adding perturbation of the above pattern (a) to the existing policy workflow. At this time, by changing a condition branch added with a perturbation and/or changing a threshold of the same condition branch, variation of the perturbation is increased and multiple types of policy candidates are generated.
In addition, when generating a policy candidate by adding a perturbation of the above pattern (b), the policy candidate generating unit 102 may increase the variation of the perturbations by changing a condition branches to be reduced and generate multiple types of policy candidates. Similarly, the policy candidate generating unit 102 may generate policy candidates by adding a perturbation of the pattern (c) or the pattern (d).
For example, in the policy candidate with a policy No. 1, the threshold of the branch condition L1 is changed to 40 from 50 of that of the policy candidates corresponding to the existing policy (with a policy No. 0). In addition, in the policy candidate with a policy No. 4, a branch condition L9 is added from the policy candidates corresponding to the existing policy (with a policy No. 0).
The policy candidate generating unit 102 generates multiple candidate workflows (policy candidates) in which a perturbation (change) is added to a part of an existing workflow that defines intervening nodes (following contents) that multiple condition branches and branch destinations reach.
The information of the policy candidates generated by the policy candidate generating unit 102 may be stored in a predetermined storing region of the storing device 10d.
The policy candidate KPI predicting unit 103 predicts a KPI of each policy candidate generated by the policy candidate generating unit 102.
The policy candidate KPI predicting unit 103 first trances, for each individual, a policy workflow through multiple condition branches on the basis of the individual state data and determines an intervening node (following contents) to be reached by specifying the intervening node.
In
In the policy candidate represented by the reference symbol A in
Then, the policy candidate KPI predicting unit 103 performs various KPI prediction based on the individual state data and the intervening node allocated to the individual.
In
For example, if the KPI is a CKD new onset ratio, the policy candidate KPI predicting unit 103 inputs the state (s1, . . . , sN) and intervening nodes (T1, . . . , TM) to a predetermined predicting model (KPI predicting model) for each individual and responsively outputs a CKD new onset ratio. A predicting model may be provided for each KPI.
A predicting model may be constructed by applying a machine learning algorithm such as Bayesian model learning or deep learning to the previous machine learning data.
A predicting model may be, for example, a deep learning model (deep neural network). A neural network may be hardware circuitry, or may be a virtual network provided by means of software that connects layers virtually constructed on a computer program by the processor 10a (see
The example of these
The individual of each policy illustrated in
As a result, as illustrated in
The policy candidate KPI predicting unit 103 obtains a KPI predicted value for each of all individuals by inputting information on the state of the individual and information on the intervening node of the individual to a KPI predicting model for each policy. A KPI predicting model is prepared for each KPI.
The example illustrated in
The policy candidate KPI predicting unit 103 calculates the KPI #1 predicted value of a policy candidate by calculating the average value of KPI #1 predicted value for all individuals calculated for the same policy candidate. Similarly, the policy candidate KPI predicting unit 103 calculates the KPI #2 predicted value of a policy candidate by calculating the average value of KPI #2 predicted value for all individuals calculated for the same policy candidate.
As illustrated in
Further, the policy candidate KPI predicting unit 103 obtains, for every policy candidate, a difference between a KPI predicted value of an existing policy and a KPI predicted value of the policy candidate, and obtains a fluctuation vector of each perturbation. A vector representing a difference between a KPI predicted value of an existing policy and a KPI predicted value of each policy candidate may be referred to as a difference vector.
The policy candidate KPI predicting unit 103 calculates a KPI fluctuation vector (difference between a KPI predicted value of the existing workflow and a KPI predicted value of each of the multiple candidate workflows) of the candidate workflow (policy candidates) from the existing workflow.
The policy candidate KPI predicting unit 103 stores information of the generated fluctuation vector of the KPI predicted value of each policy candidate into a predetermined storing region of the storing device 10d.
The synthesized policy provisional KPI predicting unit 104 generates a synthesized policy based on the respective KPI predicted value of each policy candidates calculated by the policy candidate KPI predicting unit 103. The synthesized policy provisional KPI predicting unit 104 generates multiple synthesized policies by combining two or more of the multiple KPI fluctuation vectors.
If generating multiple policy candidates to be synthesized that occur perturbations having overlapping influence ranges, the synthesized policy provisional KPI predicting unit 104 predicts a KPI, setting a higher weight to a policy candidate occurring a perturbation at the upper side when a KPI fluctuation vector is synthesized.
If a selected route in a policy workflow contains two or more perturbations, these perturbations have overlapping influence ranges.
As illustrated in
In
The target threshold #1 is set for KPI #1 and the target threshold #2 is set for KPI #2. A region satisfying both of the target thresholds #1 and #2 may be referred to as a KPI target goal region.
The synthesized policy provisional KPI predicting unit 104 sets the weight (0 or 1 in the present embodiment) for each policy candidate to generate synthesized policy satisfying both of the target thresholds #1 and #2, in other words, being included in KPI target goal region.
The present workflow generating apparatus 1 generates a synthesized policy that satisfies all KPI targets by weighting the respective perturbations and adding the weighted perturbations to the existing policy. Therefore, the synthesized policy provisional KPI predicting unit 104 calculates a provisional value of the KPI prediction of the synthesized policy.
In the present workflow generating apparatus 1, the following Expression (1) represents a KPI provisional predicted value Y of the synthesized policy obtained by the weighting and adding of an existing policy and the fluctuation vectors.
Y=X
0
+W
1
×ΔX
1
+ . . . +W
N
×ΔX
N (1)
Here, X0 is a KPI predicted value of the existing policy. W1 is the weight of the perturbation 1. ΔX1 is the fluctuation vector of the perturbation 1. WN is the weight of the perturbation N. ΔXN is perturbation N fluctuation vector.
The weight combination listing unit 201 generates combination pattern of the weights W1 to WN to be used in the KPI provisional predicted value Y of the synthesized policy represented by the above Expression (1). That is, the weight combination listing unit 201 generates a combination pattern of weights to be applied to the perturbations. The combination pattern of weights may be referred to as a weight pattern.
In the example of
For example, a weight pattern 1 indicates that the perturbation 1 and the perturbation 4 are used for a synthesized policy, but the perturbation 2 and the perturbation 3 are not used. For example, a weight pattern 4 indicates that only the perturbation 4 is used for a synthesized policy, but the perturbations 1 to 3 are not used.
The multiple types of combination patterns of weights illustrated in
As described above with reference to
In contrast to the above, as described above with reference to
As a solution to the above, in providing multiple perturbations to an existing policy workflow when a synthesized policy is generated, if the influence ranges of the multiple perturbations overlap, the weight adjusting parameter calculating unit 202 sets parameters (weight adjusting parameters) for the above parameters on the basis of the relationship among the perturbations of the existing policy workflow.
The weight adjusting parameter calculating unit 202 sets the weight adjusting parameter such that a weight of an upper perturbation (at an upper point) of a policy workflow exerts a higher influence than a weight of a lower perturbation (at a lower point).
The weight adjusting parameter calculating unit 202 determines a weight adjusting parameter N) for a perturbation i among the multiple perturbations in the following steps (1) to (6).
In
In
The weight adjusting parameter calculating unit 202 may generate an overlap presence/absence matrix on the basis the policy workflow using any known method, and the description of the method is omitted here.
The weight adjusting parameter calculating unit 202 extracts a submatrix P1 corresponding to the perturbation group P0 with reference to the overlap presence/absence matrix. In
In
In
When a flow depth of the target perturbation i is represented by x(i), the attenuation function for obtaining a weight adjusting parameter η is expressed by, for example, the following Expression (2).
η(i)=exp{−ax(i)} (2)
In the above Expression (2), the symbol “a” is a constant.
Furthermore, as the final value η′(i), the weight adjusting parameter calculating unit 202 performs normalization represented by the following Expression (3) such that the average value of all the parameter r′(i) becomes 1.
In the above Expression (3), the symbol M is the number of perturbations included in the target perturbation group P0. In the example indicated by the reference symbol A in
In
In
The KPI provisional predicting value calculating unit 203 calculates, for each KPI, a KPI provisional predicted value of a synthesized policy.
The KPI provisional predicting value calculating unit 203 calculates a KPI provisional predicted value Y(k) for each KPI on the basis of the following Expression (4).
Y
(k)
=X
0
(k)+η1W1×ΔX1(k)+ . . . +ηNWN×ΔXN(k) (4)
The symbol k is a value for specifying any KPI among L types of KPI, and the symbol k is a natural number equal to or greater than one. The symbol η1 is an adjusting parameter of the perturbation 1 and the symbol ηN is an adjusting parameter of the perturbation N.
When multiple perturbations (changes) are included in one selected route in the candidate workflow, the synthesized policy provisional KPI predicting unit 104 calculates a KPI predicted value (KPI Provisional predicted value) of each of the multiple synthesized policy by reflecting the weight parameter η set such that the weight (W) of an upper perturbation in the selected route is higher than the weight (W) of a lower perturbation in the selected route.
The synthesized policy provisional KPI predicting unit 104 calculates the KPI provisional predicted value (KPI predicted value) of each of the multiple synthesized policy by weighted addition of KPI predicted value of the existing workflow (existing policy) and the KPI fluctuation vectors (differences between KPI predicted value of the existing workflow and KPI predicted values of the respective candidate workflows).
The combination determining unit 204 calculates (determines) a weight combination (W1, . . . , WN) by applying a combination optimization algorithm or the like such that all the KPI predicted values Y after the weight adjustment by weight adjusting parameters η calculated by the KPI provisional predicting value calculating unit 203 satisfy the respective KPI targets.
The combination determining unit 204 checks whether provisional predicted values Y of the KPIs of all types of synthesized policy are less than the target values Z of the KPIs, that is, whether all the KPIs satisfy the KPI targets.
In the example of
Furthermore, the combination determining unit 204 checks, for example, whether the provisional predicted value Y(L) of a KPI #L of a synthesized policy is less than the target Z(L) of the KPI #1, (i.e., Y(L)<Z(L)).
Then, the combination determining unit 204 determines a weight combination pattern that can make all the KPI satisfy the respective KPI targets as an optimum weight combination pattern for generating a synthesized policy that can achieve the KPI target.
The combination determining unit 204 may select multiple weight combination patterns as the optimal weight combination pattern for generating a synthesized policy that can achieve the KPI target.
The synthesized policy actual KPI predicting unit 105 generates a synthesized policy by adding multiple perturbations to the existing policy on the basis of the weight combination pattern determined by the combination determining unit 204, and performs actual KPI prediction of the synthesized policy.
Then, the synthesized policy actual KPI predicting unit 105 checks whether all the KPI predicted results satisfy the respective target conditions (KPI targets) in the actual KPI prediction of the synthesized policy.
If a KPI prediction result that does not satisfy target condition is present as a result of the checking whether all the KPI predicted results satisfy the respective target conditions (KPI targets) of the synthesized policy, the synthesized policy actual KPI predicting unit 105 changes the manner of providing the perturbations and causes the policy candidate generating unit 102 to regenerate a synthesized policy.
As a manner of providing perturbations at this time, the synthesized policy actual KPI predicting unit 105 may provide perturbations starting from a synthesized policy close to the target condition.
For example, the synthesized policy actual KPI predicting unit 105 obtains a difference between KPI predicted value and the target value for every synthesized policy. When a synthesized policy has multiple KPIs, this difference is obtained for each KPI, and the average value of the differences is obtained.
Then, a synthesized policy having the average of the differences is smaller than a predetermined threshold or that having the smallest average may be specified as the synthesized policy.
The synthesized policy actual KPI predicting unit 105 generates new policy candidates by reproviding multiple new patterns of perturbations (see
If confirming that all the KPI prediction results satisfy the target conditions (KPI targets) as a result of the checking whether all the KPI predicted results satisfy the respective target conditions (KPI targets) of the synthesized policy, the synthesized policy actual KPI predicting unit 105 determines the synthesized policy to be a remedial policy and ends the process.
The output controlling unit 106 outputs information of the remedial policies determined by the synthesized policy provisional KPI predicting unit 104. The output controlling unit 106 may generate the outputted information including the information of the remedial policies and present the outputted information to the user (policy developer). The outputted information is information obtained by visualizing the information of the remedial policies. The output controlling unit 106 may include the information of the remedial policy in the outputted information.
For example, the output controlling unit 106 may output the outputted information to the monitor or the like via the graphic processing device 10b.
The outputted information illustrated in
In the information of the existing policy and the information of the remedial policy, the respective policy workflows (see the reference signs P02, and P12) are illustrated, and KPI information (see the reference signs P03, and P13) in which the target value and the predicted value of a KPI are associated is illustrated.
The user can compare and confirm the existing policy and the remedial policy.
In the policy workflow of the remedial policy, the visibility may be enhanced by a marker or changing the display color on a changed part (perturbation) from the policy workflow of the existing policy (see the reference signs P15, and P16).
Alternatively, in the KPI information of the existing policy, the visibility may be enhanced by changing the font and/or the display color of the value of KPI below the target value (see the reference signs P04). In addition, the visibility may be enhanced by a marker or changing the display color on a changed part that indicates a predicted value of a KPI below the target value in the existing policy has improved in KPI information of the remedial policy (refer to the reference sign P14).
The output controlling unit 106 outputs the workflow of synthesized policy whose KPI provisional predicted value (KPI predicted value) satisfies the target value.
(C) Operation:
An outline of a process in the workflow generating apparatus 1 according to an example of an embodiment configured as described above will now be described with reference to a flow chart (Steps S01 to S05) illustrated in
In Step S01, the existing policy obtaining unit 101 obtains an existing policy.
In Step S02, the policy candidate generating unit 102 generates multiple policy candidates by adding a perturbations to the existing policy workflow.
In Step S03, the policy candidate KPI predicting unit 103 predicts a KPI of each policy candidate generated by the policy candidate generating unit 102. Further, the policy candidate KPI predicting unit 103 obtains, for all the policy candidates, a difference between a KPI predicted value of an existing policy and a KPI predicted value of policy candidate, and obtains a fluctuation vector of each perturbation.
In Step S04, the synthesized policy provisional KPI predicting unit 104 generates a synthesized policy by synthesizing fluctuation vectors of the policy candidates on the basis of the respective KPI predicted values of the policy candidates calculated by the policy candidate KPI predicting unit 103.
In addition, the synthesized policy provisional KPI predicting unit 104 sets a weight adjusting parameters based on the relationship between perturbations on the existing policy workflow, and calculating a KPI provisional predicted value of each synthesized policy by reflecting the calculated weight adjusting parameters. If the influence ranges of the perturbations occur in generating multiple policy candidates to be synthesized, the synthesized policy provisional KPI predicting unit 104 predicts a KPI, setting a higher weight to a policy candidate occurring a perturbation at the upper side when a KPI fluctuation vectors are synthesized.
In Step S05, the synthesized policy actual KPI predicting unit 105 generates a synthesized policy by adding multiple perturbations to the existing policy on the basis of the weight combination pattern determined by the combination determining unit 204, and performs actual KPI prediction of the synthesized policy.
A synthesized policy that all KPI prediction results satisfy the target conditions (KPI targets) are adopted as a remedial policy.
Next, a detailed process in the workflow generating apparatus 1 according to an example of the embodiment will now be described with reference to a flow chart (Steps S1-S7 and S41-S44) illustrated in
In Step S1, the existing policy obtaining unit 101 obtains an existing policy.
In Step S2, the policy candidate generating unit 102 generates multiple policy candidates by adding perturbations to the existing policy workflow.
In Step S3, the policy candidate KPI predicting unit 103 predicts a KPI of each policy candidate generated by the policy candidate generating unit 102. Further, the policy candidate KPI predicting unit 103 obtains, for all the policy candidates, a difference between a KPI predicted value of an existing policy and a KPI predicted value of each policy candidate, and obtains a fluctuation vector of each perturbation.
The process for predicting the provisional KPIs of the synthesized policy in Step S4 includes the process of Steps S41 to S44.
In Step S41, the weight combination listing unit 201 generates multiple weight combination patterns to be applied to the perturbations.
In Step S42, in providing multiple perturbations to an existing policy workflow when a synthesized policy is generated, if the influence ranges of the perturbations overlap, the weight adjusting parameter calculating unit 202 sets parameters for the above weights on the basis of the relationship among the perturbations of the existing policy workflow.
In step S43 of steps, the KPI provisional predicting value calculating unit 203 calculates, for each KPI, a KPI provisional predicted value of the synthesized policy.
In S44 of steps, the combination determining unit 204 calculates (determines) a weight combination such that all the KPI predicted values after the weight adjustment by weight adjusting parameters calculated by the KPI provisional predicting value calculating unit 203 satisfy the respective KPI targets.
After that, in Step S5, the synthesized policy actual KPI predicting unit 105 generates a synthesized policy by adding multiple perturbations to the existing policy on the basis of the weight combination pattern determined by the combination determining unit 204, and performs actual KPI prediction of the synthesized policy.
In Step S6, the synthesized policy actual KPI predicting unit 105 checks whether all the KPI predicted results satisfy the respective target conditions (KPI targets) in the actual KPI prediction of the synthesized policy.
If a KPI prediction result that does not satisfy the target condition is present (see No route of Step S6), the process returns to Step S2 that changes the manner of providing the perturbations and causes the policy candidate generating unit 102 to regenerate a synthesized policy.
On the other hand, if all the KPI prediction results satisfy the target conditions (KPI target) as a result of the checking in Step S6 (see YES route in Step S6), the process proceeds to Step S7.
In Step S7 of steps, the output controlling unit 106 outputs the remedial policy determined by the synthesized policy provisional KPI predicting unit 104. Then, the process ends.
(D) Effect:
As described above, according to the workflow generating apparatus 1 as an example of the embodiment, if the influence ranges of the perturbations overlap in generating multiple policy candidates to be synthesized, the synthesized policy provisional KPI predicting unit 104 predicts a KPI, setting a higher weight to a policy candidate occurring a perturbation at the upper side when a KPI fluctuation vector is synthesized.
Considering that a lower perturbation is affected by an upper perturbation in a policy workflow as the above makes it possible to predict more accurate KPIs.
In addition, the weight adjusting parameter calculating unit 202 sets a higher value to a weight adjusting parameter η of a shallower (upper) flow depth, and a lower value to a weight adjusting parameter η of a deeper (lower) flow depth by using an attenuation function. This makes it possible to set a higher weight to a policy candidate occurring a perturbation at the upper side when a KPI fluctuation vector is synthesized.
(E) Miscellaneous:
The disclosed techniques are not limited to the embodiment described above, and may be variously modified without departing from the scope of the present embodiment. The respective configurations and processes of the present embodiment can be selected, omitted, and combined according to the requirement.
For example, as illustrated in
In the above-described embodiment, the weight W used in KPI provisional predicted value Y of synthesized policy is either 0 or 1, but the present invention is not limited thereto and may be a value other than 0 and 1.
Further, in the above-described embodiment, the method of adding a perturbation to the policy workflow is not limited to that illustrated in
In addition, those ordinary skilled in the art can carry out and manufacture of the present embodiments with reference to this disclosure.
According to an embodiment, a workflow that achieves a higher KPI based on an existing workflow can be generated.
Throughout the descriptions, the indefinite article “a” or “an”, or adjective “one” does not exclude a plurality.
All examples and conditional language recited herein are intended for the pedagogical purposes of aiding the reader in understanding the invention and the concepts contributed by the inventor to further the art, and are not to be construed limitations to such specifically recited examples and conditions, nor does the organization of such examples in the specification relate to a showing of the superiority and inferiority of the invention. Although one or more embodiments of the present inventions have been described in detail, it should be understood that the various changes, substitutions, and alterations could be made hereto without departing from the spirit and scope of the invention.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2022-103368 | Jun 2022 | JP | national |
