WO2024029189A1

WO2024029189A1 - Development support system

Info

Publication number: WO2024029189A1
Application number: PCT/JP2023/020992
Authority: WO
Inventors: 光博木谷; マンイウーチャウ; 智圓羅; 壮希櫻井; 尊文鈴木
Original assignee: 日立Astemo株式会社
Priority date: 2022-08-03
Filing date: 2023-06-06
Publication date: 2024-02-08
Also published as: JP2024021216A

Abstract

An example of a development support system according to the present invention comprises: a storage unit that stores at least update history data of a first component and current data of a second component of a developed product; an evaluation prediction unit that predicts an evaluation value for the development achievement level of each component; a difference extraction unit that extracts the difference between current data of the first component and pre-update data, which is one set of data in the update history data; and a data modification unit that generates modified data by modifying the current data of the second component on the basis of the output of the difference extraction unit. The evaluation prediction unit calculates a first variation value for the evaluation value on the basis of the current data and the pre-update data of the first component and a second variation value for the evaluation value on the basis of the current data and the modified data of the second component.

Description

Development support system

The present invention relates to a development support system that predicts how much a certain input variable will contribute to the outcome of a project, such as software development.

In general, in software development sites, the project goes through multiple upgrades until it is completed, that is, before the final version of the software is released. In recent years, a concept called KPI (Key Performance Index) has been introduced as a standard for determining how close each version is to the final goal. KPI is also called "Key Performance Indicator" and is generally a quantitative measurement standard set to manage the process of achieving the final goal of a project. Specific examples of KPI include CPU load, ROM consumption, RAM consumption, number of bugs, development man-hours, software reuse rate, and the like.

Techniques for predicting this KPI using software architecture design data metrics have also been proposed in recent years. Here, the software architecture design data metrics define standards for quantitatively evaluating the software architecture (software structure, functional/non-functional requirements, etc.) that is a development target in a certain project. Specific examples include the number of modules per core, number of dependencies/amount of data, direction of dependencies, degree of cyclomatic complexity, etc. Other examples include measuring system scale, understanding the sensitivity of the system to changes, Various criteria are adopted from the viewpoints of understanding the degree of connection between elements, the degree of coupling between elements, and the degree of circulation of elements. Note that hereinafter, software architecture design may be simply referred to as software architecture design, and software architecture design data metrics may simply be referred to as software architecture design data metrics.

In the case of a software version change, in addition to the software architecture design data, information regarding the hardware and project may also be changed. Furthermore, as software development projects have become larger and more complex in recent years, the number of variables required for KPI prediction continues to increase. Therefore, it is necessary to establish a method for calculating the absolute amount of influence of each input variable on the predicted results of KPI, and a method for calculating the influence of software architecture design changes.

Patent Document 1 discloses a technique for randomly changing inputs within a value range and specifying inputs (features) for achieving an objective regarding a target.

In addition, Patent Document 2 discloses that a (second) project success probability is calculated using a prediction model from a project parameter set and a scheduled end date, and a new parameter set and a scheduled end date are calculated when there is a change in the project. discloses a technique for similarly calculating the (first) project success probability and estimating the risk due to project change from the difference between the first and second success probabilities.

Japanese Patent Application Publication No. 2020-119085 JP2020-091843A

However, the technology disclosed in Patent Document 1 does not take into consideration the handling of data that is version-controlled and continuous over time, such as software design data, and does not take into account the predicted value of KPI impact on architectural design changes. (absolute amount) cannot be calculated. The same applies to Patent Document 2.

The present invention has been made in view of the above points, and when the software architecture design is changed during project execution, the degree of influence of other input variables on KPI is predicted using a prediction model, and the software architecture can be changed at an early stage. The purpose is to provide a development support system that can verify the validity of a design.

An example of the development support system according to the present invention includes a storage unit that stores at least update history data of a first component of a development product and current data of a second component, and an evaluation value for the degree of development achievement for each component. an evaluation prediction unit that predicts a difference between the current data of the first component and pre-update data that is one of the update history data; , a data change unit that generates changed data obtained by changing the current data of the second component, and the evaluation prediction unit changes the first evaluation value based on the current data of the first component and the pre-update data. A fluctuation value is calculated, and a second fluctuation value of the evaluation value is calculated based on the current data and the changed data of the second component.

According to the present invention, when the soft architecture design is changed during the execution of a project, it is possible to predict the degree of influence of other input variables on KPI using a prediction model, and to verify the validity of the soft architecture design at an early stage. become.
Further features related to the invention will become apparent from the description herein and the accompanying drawings. Further, problems, configurations, and effects other than those described above will be made clear by the description of the following examples.

The figure which shows the outline of the method of predicting KPI using a prediction model. FIG. 1 is a block diagram showing the functional configuration of a development support system according to an embodiment of the present invention. FIG. 3 is a diagram showing relationships between functional units in processing executed by the development support system. FIG. 3 is a diagram for explaining a method of calculating a comparison of variation ratios between variables with respect to KPI when a development support system is applied. A flowchart showing processing executed by the development support system.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a diagram showing an overview of a method for predicting KPIs by inputting softarchitecture design data and KPI-related data other than the softarchitecture design data into a KPI prediction model. The software architecture design data is data indicating the structural requirements of the software to be developed, and includes the number of modules by core, number of dependencies/data amount, direction of dependencies, degree of cyclomatic complexity, etc. as described above. KPI-related data does not refer to data related to the software being developed itself, but refers to other data related to the development project, including project management data related to the project itself such as similarities with past projects and customer requests, and information about microcontrollers. This includes hardware design data such as specifications. Specifically, the data types are as shown by variable names X2 to X8 in FIG. 1, for example.

Then, these data are input to a KPI prediction model using AI that has been trained on data showing the relationship between past input variables and actually calculated KPIs, and KPIs are predicted. In FIG. 1, examples of KPIs include Software productivity [LoC/hour] (number of source code lines generated per unit time: indicates productivity), CPU workload [%] (CPU load rate), and Software bugs [ Number/LoC] (number of bugs per line of source code).

As shown in Figure 1, it can be seen that each KPI changes each time the version of the software to be developed is revised. These results are fed back to the AI of the KPI prediction model, improving its prediction accuracy.

However, as a matter of course, each software architecture design data and KPI-related data will also be changed as the software version is revised. As mentioned above, the number of variables that need to be handled has increased due to the increasing complexity of software development projects in recent years, and while it is possible to predict the impact of changes in all variables on KPIs to some extent, changes in each variable are It is very difficult to grasp the extent to which changes are being influenced.

The present invention has been made based on the above background, and FIG. 2 shows a functional configuration block diagram of a development support system 1 according to an embodiment of the present invention. In addition, the development support system 1 may be a computer equipped with a memory and a processor, for example, as a hardware configuration, or may be a cloud installed on a server.

The development support system 1 includes a normalization processing section 11, a difference extraction section 12, a scale adjustment section 13, a data modification section 14, an evaluation prediction section 15, a fluctuation ratio calculation section 16, a fluctuation ratio comparison section 17, and an evaluation result output section 18. Equipped with Details of these functions will be described later. The development support system 1 also includes a storage unit 100.

The storage unit 100 stores software architecture design data 101 and KPI-related data 102, and the KPI-related data 102 further includes hardware design data 1021 and project management data 1022. These software architecture design data 101 and KPI related data 102 are stored in association with each software version.

The development support system 1 is also connected to an external server 3 via a communication path 2 connected to a communication interface (not shown) of the development support system 1. The communication path 2 may physically include a plurality of communication buses, and the standards of each communication bus may be the same or different. The external server 3 sends and receives messages to and from the development support system 1 via the communication path 2.

The functional block diagram shown in FIG. 2 is an example, and the functional units and names are not limited to this. For example, the functions realized by the scale adjustment section 13 in this embodiment may be realized by other functional sections shown in FIG. 1, or may be realized by a functional section not shown in FIG.

FIG. 3 is a diagram showing the relationship between each functional unit in the processing executed by the development support system. Here, as an example, processing when the software version is updated from v2.0 to v3.0 will be described with reference to FIG. First, the normalization processing unit 11 extracts the current data (data of software version v3.0) and the data before update (data of software version v2.0) of the software architecture design data 101 from the storage unit 100, and normalizes them. processing. The normalization mentioned here can be in any format. For example, linear transformation or affine transformation may be used to set the maximum value of each data to 1 and the minimum value to 0.

Subsequently, the difference extraction unit 12 extracts the difference between the current data processed by the normalization processing unit 11 and the pre-update data. This difference is a difference normalized to 0 to 1, so it takes a value of -1 to 1.

The extracted difference is passed to the scale adjustment section 13. At the same time, or prior to this, the pre-data change metrics value calculation unit 141 of the data change unit 14 refers to the KPI-related data 102 in the storage unit 100, extracts current data (data of software version v3.0), Measure from a preset viewpoint and calculate a metric value. At this time, a plurality of pieces of data may be extracted from the KPI-related data and collectively calculated as a metrics value. The calculated metrics value is passed to the scale adjustment section 13.

The scale adjustment unit 13 associates the difference received from the difference extraction unit 12 with the metrics value regarding the current data of the KPI-related data 102 received from the pre-data change metric value calculation unit 141 and passes it to the post-data change metric value calculation unit 142. .

Based on the received data, the post-data change metrics value calculation unit 142 calculates the metrics value of the current KPI-related data with the normalized difference between the software architecture design data metrics values before and after the update, extracted by the difference extraction unit 12. Change to a value with the same amount of difference. Specifically, for example, if the normalized difference in the soft arch design data metrics value is 0.5, that is, from the soft arch design data metrics value before the update to the soft arch design data metrics value after the update, Assuming that the value has increased by 50% of the maximum value, the post-data change metrics value calculation unit 142 subtracts 0.5 from the normalized value of the current KPI-related data metrics value, and further What is quantified using the standard before normalization is calculated as the changed KPI-related data metrics value. These values are passed to the evaluation prediction unit 15.

The pre-data change predicted evaluation value calculation unit 151 of the evaluation prediction unit 15 predicts the evaluation of the current (before data change) KPI-related data metrics value received from the pre-data change metrics value calculation unit 141 with respect to the KPI. Similarly, the post-data change evaluation predicted value calculation unit 152 predicts the evaluation of the post-data change KPI-related data metrics received from the post-data change metrics value calculation unit 142 with respect to the KPI. These calculated values are passed to the fluctuation ratio calculation section 16.

The variation ratio calculation unit 16 compares the two predicted values received from the evaluation prediction unit 15 and calculates the variation ratio. In other words, the change from the KPI-related data metric value that was changed based on the difference between the normalized software architecture design data metric values before and after the update to the current KPI-related data metric value is different from the current KPI-related data metric value. Calculate the percentage of The calculated ratio is passed to the fluctuation ratio comparison section 17.

After receiving the fluctuation ratios for all KPI-related data metrics values from the fluctuation ratio calculation unit 16, the fluctuation ratio comparison unit 17 compares the fluctuation ratios among all variables, including the fluctuation ratios of software architecture design data metrics values. Compare. The calculated result is output as the input variable influence degree evaluation result 19 from the evaluation result output unit 18 to an external administrator or the like.

FIG. 4 is a diagram visually showing the process described in FIG. 3, and is a diagram for explaining a method of calculating a comparison of variation ratios between variables with respect to KPI when the development support system is applied.

First, the values before and after updating of any software architecture design data metrics represented by the variable name X1 are input to the AI that constitutes the KPI prediction model, and the KPI evaluation prediction for each is calculated. As an example, in Figure 4, KPI is the number of bugs, and the predicted number of bugs for the software architecture design data metrics value in the current version (after update) is 20, and the number of bugs predicted for the software architecture design data metrics value in the old version (before update) is 20. The expected number of bugs for the value is 10.

From this, the ratio of the KPI fluctuation value (10) before and after the update to the number of bugs (20) predicted for the software architecture design data metrics value in the current version is 50%, so it can be seen that In this step, the variation ratio calculation unit 16 calculates 50% as the variation ratio of the software architecture design data metrics value represented by the variable name X1 with respect to the KPI.

In parallel with this, the KPI is also predicted for the KPI-related data metric (past project similarity metric in FIG. 4) represented by the variable name X2, and the rate of change is calculated. As described above, the KPI-related data metrics values are changed to match the normalized difference between the softarchitecture design data metrics values before and after the update. In Figure 4, the normalized value of the current past project similarity metric is 0.7, and the above-mentioned difference between the normalized software architecture design data metric values before and after the update is -0.02. , the normalized value of the changed past project similarity metric is set to 0.72.

Then, these values are input into the KPI prediction model, and a predicted KPI evaluation for each is calculated. In the example of Figure 4, the number of bugs predicted for the current past project similarity metric value is 20, and the number of bugs predicted for the past project similarity metric value after data change is 18. .

From this, the variation value of the past project similarity metric value before and after the change is 2, and the proportion of the predicted number of bugs 20 for the current past project similarity metric value is 10%, so the figure In step 4, the variation rate calculation unit 16 calculates 10% as the variation rate of the past project similarity metric value with respect to the KPI.

The above process is similarly performed for the remaining KPI-related data metrics represented by variable names X3 to X5. As a result, the variation ratio calculation unit 16 calculates 15%, 8%, and 30% as the variation ratios of the KPI-related data metrics represented by the variable names X3 to X5 with respect to the KPI, respectively.

Then, the variation rate comparison unit 17 compares the variation rates between variables based on the calculation result of the variation rate calculation unit 16 described above. This can be obtained, for example, by leveling out each variation ratio so that the total becomes 100%. From this result, when updating the software version, the influence that the software architecture design data metrics value has on the fluctuation of KPI is 44% of the total, and for example, the complexity metric value of the request represented by variable name It can be seen that the influence on fluctuations is 13% of the total.

Finally, the flow of processing executed by the development support system 1 will be explained using the flowchart of FIG. 5.

First, in step S501, arbitrary data is extracted from the software architecture design data and KPI-related data for each version, and is digitized (metric value calculation) by being evaluated based on a predetermined standard. This process may be executed, for example, by the data change unit 14 or by a CPU (not shown) included in the development support system.

Next, in step S502, the software architecture design data metrics change rate between versions is calculated. Specifically, this is executed by the normalization process (step S5021) by the normalization processing unit 11 and the difference extraction process (step S5022) by the difference extraction unit 12, which have already been described.

Subsequently, in step S503, the scale adjustment unit 13 performs scale adjustment based on the current (before change) KPI-related data metrics value and the software architecture design data metrics value difference between versions, and in step S504, the post-data change metrics value The data is changed by the calculation unit 142.

In step S505, the changed KPI-related data metrics value is input to the evaluation prediction unit 15, and a KPI predicted value is calculated.

In step S507, it is determined whether the above processing has been executed for all input variables, and the processing of steps S504 to S505 is repeated until it has been executed for all input variables.

Also, in parallel with or in advance of the above processing, in step S506, KPI predicted values of the current software architecture design data metrics values and KPI related data metrics values are calculated.

In step S508, the fluctuation ratio calculation unit 16 calculates the fluctuation ratio from the KPI predicted value of the KPI-related data metrics values before and after the change.

Finally, in step S509, the variation rate comparison unit 17 compares the variation rate of the KPI predicted value between the input variables, and calculates the result as the input variable influence degree evaluation result 19.

As explained above, according to this embodiment, the KPI-related data metrics values are adjusted to have the same fluctuation range before and after updating the known software architecture design data metrics values, and the KPI prediction model is used to Predicting KPIs. This makes it possible to evaluate the degree of influence of each variable on the KPI, including the softarchitecture design data metrics values. Therefore, using the evaluation results, it becomes possible to quickly evaluate the validity of the software architecture design.

According to the embodiments of the present invention described above, the following effects are achieved.
(1) An example of the development support system according to the present invention includes a storage unit that stores at least update history data of a first component of a development product and current data of a second component; , an evaluation prediction unit that predicts an evaluation value for the degree of development performance; a difference extraction unit that extracts a difference between the current data of the first component and pre-update data that is one of the update history data; a data modification unit that generates changed data obtained by changing the current data of the second component based on the output of the extraction unit; A first variation value of the evaluation value is calculated based on the current data and the changed data of the second component, and a second variation value of the evaluation value is calculated based on the current data and the changed data of the second component.

With the above configuration, if the soft architecture design is changed during project implementation, it is possible to predict the influence of other input variables on KPI using a prediction model and verify the validity of the soft architecture design at an early stage. .

(2) Further comprising a normalization processing unit that normalizes the current data of the first component and the pre-update data, and the difference extraction unit is configured to normalize the current data of the first component and the pre-update data. After extracting the difference, the data modification section modifies the current data of the second component so as to correspond to the normalized difference extracted by the difference extraction section to generate modified data. As a result, even if the granularity or unit of data differs between constituent elements, it becomes possible to easily compare data by normalizing and making it dimensionless.

(3) The apparatus further includes a variation ratio calculation unit that calculates a variation ratio of each of the first variation value and the second variation value to the degree of development performance. This makes it possible to compare the degree of influence on the development results between the variable values, and it becomes possible to easily determine which component has a greater influence on the development results.

(4) The evaluation prediction unit further includes a fluctuation ratio comparison unit that predicts evaluation values of the plurality of second components and compares the ratio of each fluctuation value between the first component and the plurality of second components. Be prepared. As a result, in addition to the effect (3), when there are multiple constituent elements, it becomes possible to rank and understand the magnitude of their influence.

(5) The device further includes an output section that outputs the first variation value and the second variation value. This makes it possible to output the evaluation results to a display device or the like that can be viewed by an external administrator or the like, improving the convenience of information management for the administrator or the like.

(6) The development product is software, the first component is a component included in the software itself, and the second component is a component related to the software development environment or a component related to the hardware for executing the software. Contains at least one of the following. The present invention can be suitably applied in such a software development environment.

Note that the present invention is not limited to the above embodiments, and various modifications are possible. For example, the above-mentioned embodiments have been described in detail to explain the present invention in an easy-to-understand manner, and the present invention is not necessarily limited to embodiments having all the configurations described. Furthermore, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment. Further, it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Further, it is possible to delete a part of the configuration of each embodiment, or to add or replace other configurations.

1 Development support system, 11 Normalization processing unit, 12 Difference extraction unit, 13 Scale adjustment unit, 14 Data modification unit, 15 Evaluation prediction unit, 16 Change rate calculation unit, 17 Change rate comparison unit, 18 Evaluation result output unit (output section), 100 storage section, 101 software architecture design data, 102 KPI related data

Claims

A development support system that supports evaluation of development products,
a storage unit that stores at least update history data of a first component and current data of a second component of the development product;
an evaluation prediction unit that predicts an evaluation value for the degree of development achievement for each component of the development product; current data of the first component; and pre-update data that is one of the update history data; a difference extraction unit that extracts the difference between
a data changing unit that generates changed data obtained by changing the current data of the second component based on the output of the difference extracting unit;
The evaluation prediction unit calculates a first variation value of the evaluation value based on the current data of the first component and the pre-update data, and calculates a first variation value of the evaluation value based on the current data of the second component and the post-change data. calculating a second variation value of the evaluation value based on
A development support system characterized by:
The development support system according to claim 1,
further comprising a normalization processing unit that normalizes the current data of the first component and the pre-update data,
The difference extraction unit extracts a difference between the normalized current data of the first component and the pre-update data,
The data changing unit changes the current data of the second component to correspond to the normalized difference extracted by the difference extracting unit to generate the changed data.
A development support system characterized by:
The development support system according to claim 1,
further comprising a variation ratio calculation unit that calculates a variation ratio of each of the first variation value and the second variation value with respect to the development achievement level;
A development support system characterized by:
The development support system according to claim 3,
The evaluation prediction unit predicts evaluation values of the plurality of second components,
further comprising a variation ratio comparison unit that compares the ratio of the variation values between the first component and the plurality of second components,
A development support system characterized by:
The development support system according to claim 1,
further comprising an output unit that outputs the first variation value and the second variation value,
A development support system characterized by:
The development support system according to claim 1,
The development product is software;
The first component is a component included in the software itself,
The second component includes at least either a component related to a development environment for the software or a component related to hardware for executing the software.
A development support system characterized by: