Information processing device and quality evaluation method

Abstract

To provide an information processing device and a quality evaluation method that can perform further appropriate evaluation of validity of each development process.SOLUTION: An information processing device according to one aspect of the present invention includes: a creation unit that creates multiple prediction model candidates using, as data for learning, first quality control data with known evaluation results; a determination unit that determines prediction accuracy of the multiple prediction model candidates using, as input information, second quality control data with known evaluation results; an extraction unit that extracts, as a representative prediction model, at least one prediction model candidate from the multiple prediction model candidates based on the determination results of the prediction accuracy of the multiple prediction model candidates; and an evaluation unit that evaluates quality of third quality control data being an evaluation target, using the representative prediction model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing apparatus and a quality evaluation method for evaluating the validity of each development process based on quality control data in each design process of system development. [Background technology]

[0002] Techniques for calculating a quality target value according to the complexity of a design have been proposed. For example, Patent Document 1 discloses a quality target value calculation device that includes a basic data extraction means that analyzes a design document and extracts basic data related to the design document, an element calculation means that calculates multiple complexity elements that are index values ​​indicating the complexity of the design document based on the basic data, a complexity calculation means that calculates the complexity of the design document using the multiple elements, and a target value calculation means that calculates a quality target value of the design document using the complexity calculated by the complexity calculation means. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2012-173829 Summary of the Invention [Problem to be solved by the invention]

[0004] In the software development process, proper review of quality control data at each stage of development is required to detect defects early. In particular, with the increasing sophistication and complexity of systems in recent years, there is a need for more accurate evaluation of the impact that quality control data has on the quality of the final deliverable.

[0005] In view of the above circumstances, an object of the present invention is to provide an information processing device and a quality evaluation method that can more appropriately evaluate the validity of each development process (design validity in the design process, program validity in the programming process, and test validity in the testing process). [Means for solving the problem]

[0006] An information processing device according to one embodiment of the present invention includes a creation unit that creates multiple prediction model candidates using first quality control data, the evaluation results of which are known, as learning data; a determination unit that determines the prediction accuracy of the multiple prediction model candidates using second quality control data, the evaluation results of which are known, as input information; an extraction unit that extracts at least one prediction model candidate from the multiple prediction model candidates as a representative prediction model based on the determination results of the prediction accuracy of the multiple prediction model candidates; and an evaluation unit that evaluates the quality of third quality control data, the evaluation target, using the representative prediction model.

[0007] The information processing device evaluates quality control data using a representative prediction model extracted from multiple candidate prediction models based on their prediction accuracy, thereby improving the reliability of evaluations based on quality control data. This allows for more accurate evaluation of the validity of each development process at the design stage.

[0008] The first quality control data may include quality data including aggregated values ​​regarding defects aggregated for each evaluation item in each development process of two or more software development projects, and final quality results. By using quality data related to two or more software development projects as a learning model, the versatility of the evaluation results is increased, thereby further improving the accuracy of the quality evaluation.

[0009] The first quality control data typically includes both quality data for which the quality result is good and quality data for which the quality result is bad, thereby improving the accuracy of the candidate prediction model to be created.

[0010] Each evaluation item of the first quality control data may be classified into multiple data groups consisting of combinations of similar evaluation items, and the creation unit may create multiple prediction model candidates corresponding to multiple types of data group patterns obtained by arbitrarily combining the multiple data groups according to the development process. This reduces the amount of data required to train each prediction model candidate, thereby shortening the time required to create a prediction model candidate.

[0011] The plurality of types of data group patterns may include an essential data group that is a data group common to each data group pattern. Since the training data for each created prediction model candidate contains at least the same set of essential data, significant variations in prediction accuracy are reduced.

[0012] The extraction unit may extract three or more candidate prediction models from the plurality of candidate prediction models as the representative prediction model, thereby improving the reliability of the quality prediction result by the evaluation unit.

[0013] The second quality control data may be the first quality control data. In this case, the determination unit may determine a degree of agreement between a prediction result obtained using the first quality control data as input information and the quality result included in the first quality control data for each of the two or more software development projects. This can improve the reliability of prediction accuracy.

[0014] The extraction unit may extract, as the representative prediction model, a prediction model candidate for which the degree of agreement of the prediction results in all software development projects is equal to or greater than a predetermined value.

[0015] A quality evaluation method according to one embodiment of the present invention is a quality evaluation method executed on an information processing device, which creates multiple candidate prediction models based on first quality control data for which the evaluation results are known, determines the prediction accuracy of the multiple candidate prediction models using second quality control data for which the evaluation results are known as input information, extracts at least one candidate prediction model from the multiple candidate prediction models as a representative prediction model based on the determination results of the prediction accuracy of the multiple candidate prediction models, and evaluates the quality of third quality control data to be evaluated using the representative prediction model. [Effects of the Invention]

[0016] According to the present invention, the validity of each development process can be evaluated more appropriately. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a block diagram showing a configuration of an information processing system including an information processing apparatus according to an embodiment of the present invention. [Figure 2] 10 is a flowchart illustrating a processing procedure executed in the information processing device. [Figure 3] FIG. 10 is a process diagram illustrating a procedure for preparing first quality control data. [Figure 4] FIG. 10 is a process diagram illustrating a procedure for preparing first quality control data. [Figure 5] FIG. 10 is a process diagram illustrating a procedure for preparing first quality control data. [Figure 6] FIG. 10 is a process diagram illustrating a procedure for preparing first quality control data. [Figure 7] FIG. 10 is a process diagram illustrating a procedure for preparing first quality control data. [Figure 8] FIG. 10 is a schematic diagram illustrating a process for creating a prediction model candidate. [Figure 9] FIG. 10 is a process diagram illustrating a method for determining the prediction accuracy of a candidate prediction model. [Figure 10] FIG. 10 is a process diagram illustrating a method for determining the prediction accuracy of a candidate prediction model. [Figure 11]FIG. 1 is a process diagram illustrating a method for extracting a representative prediction model. [Figure 12] FIG. 1 is a schematic diagram illustrating quality evaluation using a representative prediction model. DETAILED DESCRIPTION OF THE INVENTION

[0018] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0019] An information processing device according to one embodiment of the present invention is used to evaluate the suitability (validity) of each development step obtained in each development step of new software. Generally, new software development projects are divided into multiple processes, and include, for example, from the upstream side, multiple development processes including the planning process, requirements confirmation process, basic design process, detailed design process, program design process, production process, unit testing process, integration testing process, and comprehensive testing process, with the downstream development process being carried out using the deliverables from the upstream development process. However, if the deliverables from the upstream development phases contain many defects, this will have a significant impact on the evaluation of the validity of each development phase in the final development phase.In addition, if a serious defect is discovered in the final deliverable, it is necessary to consider in which development phase the defect that caused it occurred, but the further upstream the development phase in which the defect occurred, the more difficult it becomes to resolve it. Therefore, it is necessary to properly evaluate the validity of each development process to determine whether the deliverables of each development process satisfy the required specifications. The information processing device of this embodiment is used as a tool for evaluating the validity of each development process. The information processing apparatus of this embodiment will be described in detail below.

[0020] [Information Processing Systems] FIG. 1 is a block diagram showing the configuration of an information processing system 1 including an information processing device 10 according to this embodiment.

[0021] The information processing system 1 includes an information processing device 10, a database 20, and an input / output unit 30.

[0022] The information processing device 10 is configured as a computer having an internal memory such as a CPU (Central Processing Unit) and a ROM (Read Only Memory) / RAM (Random Access Memory). As will be described in detail later, the information processing device 10 has a function of generating a machine learning model for evaluating the suitability (validity) of each development process, which is the deliverable of each development process of software currently being developed, by using process quality data of software developed in the past as learning data.

[0023] The information processing device 10 has, as functional blocks of a CPU, a creation unit 11, a determination unit 12, an extraction unit 13, and an evaluation unit 14. Note that the information processing device 10 is not limited to being constructed by a single computer, and any one or more functional blocks may be constructed by separate computers.

[0024] The database 20 stores various parameters necessary for the programs and calculations executed by the information processing device 10, as well as machine learning models generated by the information processing device 10. Note that the programs and various parameters necessary for the calculations executed by the information processing device 10 may be stored in an internal memory of the information processing device 10.

[0025] The input / output unit 30 is a user interface that displays various data generated in the information processing device 10, or inputs settings for various functions and parameter changes of the information processing device 10. The input / output unit 30 is, for example, an input device having a display screen.

[0026] [Details of information processing device] 2 is a flowchart showing a processing procedure executed in the information processing device 10. Details of the information processing device 10 will be described below with reference to FIG.

[0027] (Creation Department) The creation unit 11 is a generator of a machine learning model equipped with artificial intelligence (AI). As the creation unit 11, a known predictive analysis tool can be used.

[0028] The creating unit 11 creates first quality control data for which the evaluation result is known. are used as training data to create multiple prediction model candidates (ST101).

[0029] The first quality control data includes, for example, process quality data for each development process in a past software development project and the success or failure of that development (whether or not the final product of that development meets the quality standards). The first quality control data is not limited to only those whose success or failure in the development was judged to be good (◯), but preferably also includes those whose success or failure was judged to be bad (×). This can improve the accuracy of the prediction model candidate to be created.

[0030] Here, process quality data refers to the quantified data of evaluation results for each development process. For example, process quality data for the detailed design process refers to the aggregated results of issues identified in design document reviews, while process quality data for the unit testing process and integration testing process refers to the aggregated results of bugs detected in testing.

[0031] If the process quality data to be evaluated is, for example, process quality data from the detailed design process, not only the process quality data from this detailed design process but also the process quality data from the upstream process (for example, the basic design process) will be evaluated. In other words, for example, if the validity of the process up to the third process is to be evaluated, the evaluation targets will be the first to third processes, and if the validity of the process up to the second process is to be evaluated, the evaluation targets will be the first and second processes.

[0032] Furthermore, the process quality data can be used from any software development process developed in the past, regardless of the development target or development scale.

[0033] The candidate prediction model is a machine learning model (a trained model trained using the first quality control data) that can be employed as a quality evaluation tool for evaluating the quality of the developed process quality data. As will be described later in this embodiment, multiple types of the first quality control data are used as training data to create multiple candidate prediction models with different specifications.

[0034] (Preparation of the first quality control data) 3 to 8 are process diagrams illustrating the procedure for preparing the first quality control data.

[0035] First, as shown in Fig. 3, a first summary table 101 is created by aggregating quality data for each item (quality data for each development process and final quality results) from process quality data DP1, DP2, ..., DPn for each development process of two or more past software development projects whose evaluation results are known. Note that this first summary table 101 is created by a worker.

[0036] In the first summary table 101, evaluation items are listed in the rows and project names are listed in the columns. There is no particular limit to the number of evaluation items, but here, an example will be described in which the number of items is 100 (n=100). It is preferable that each evaluation item is common to or corresponds to all project names. Examples of item names include industry, development scale, design density, review issue density, test density, defect density, etc., and each item contains the summary results of issues and bugs identified in the design review.

[0037] The "quality result" in the rightmost column of the first summary table 101 indicates the success or failure of the corresponding software development project as a whole, and is a two-stage evaluation result of good (◯) or bad (×). In the following explanation, the development process in each development project is assumed to include nine processes, from upstream to downstream: planning process, requirements confirmation process, basic design process, detailed design process, program design process, production process, unit testing process, integration testing process, and comprehensive testing process.

[0038] Next, as shown in Fig. 4, each evaluation item in the first summary table 101 is classified into similar data groups. Here, items 1 to 10 are classified as data group 0, items 11 to 30 as data group 1, items 31 to 50 as data group 2, items 51 to 70 as data group 3, and items 71 to 100 as data group 4.

[0039] Data group 0 includes, for example, essential items common to each development project, such as the type of industry and the scale of development.

[0040] Data group 1 includes, for example, items related to design / review / test / defect density, specifically, design density (number of design document pages), review issue density (number of review issues), test density (test items), defect density (number of defects), etc.

[0041] Data group 2 includes, for example, items similar to the inter-process cumulative totals and ratios of data group 1, such as the ratio of the number of design document pages to the number of review findings, the ratio of the number of tests to the number of defects, etc.

[0042] Data group 3 includes, for example, items similar to the cause occurrence process, and specifically includes causes attributable to the planning process (number of issues raised in the review of the design documents in the planning process), causes attributable to the requirements confirmation process (number of issues raised in the review of the design documents in the requirements confirmation process), causes attributable to the basic design (number of issues raised in the review of the design documents in the basic design process), causes attributable to the detailed design (number of issues raised in the review of the design documents in the detailed design process), causes attributable to the program design (number of issues raised in the review of the design documents in the program design process), causes attributable to the production (number of issues raised in the review of the program code), causes attributable to the unit testing (number of actual defects that occurred in the test), causes attributable to the integration testing (number of actual defects that occurred in the test), etc.

[0043] Data group 4 includes, for example, items similar to the in-process ratios in data group 3, such as the basic design attribution ratio (the ratio of basic design attribution to the sum of planning process attributions, requirement confirmation process attributions, and basic design attributions), the detailed design attribution ratio (the ratio of detailed design attribution to the sum of planning process attributions, requirement confirmation process attributions, basic design attributions, and detailed design attributions), the program design attribution ratio (the ratio of program design attribution to the sum of planning process attributions, requirement confirmation process attributions, basic design attributions, detailed design attributions, program design attributions, production attributions, and unit test attributions), and the production attribution ratio (the ratio of production attribution to the sum of planning process attributions, requirement confirmation process attributions, basic design attributions, detailed design attributions, program design attributions, production attributions, and unit test attributions).

[0044] The item names and data groups of the first summary table 101 are set according to the development process to be evaluated. Examples of data groups other than those mentioned above include data groups similar to the types of defect causes (design error, work error, skill shortage, specification change, pointed out error, program error, test error, etc.), and data groups similar to the intra-process ratios and inter-process cumulative totals of the above data groups.

[0045] Next, all combination patterns of the data groups are created. Here, each pattern includes data group 0 as a common required data group, and all combinations of the remaining four data groups 1 to 4 are created. As a result, the following 15 types of data group patterns are obtained, as shown in Figures 5 and 6.

[0046] Pattern 1: Data group 0 + Data group 1 Pattern 2: Data group 0 + Data group 2 Pattern 3: Data group 0 + Data group 3 Pattern 4: Data group 0 + Data group 4 Pattern 5: Data group 0 + Data group 1 + Data group 2 Pattern 6: Data group 0 + Data group 1 + Data group 3 Pattern 7: Data group 0 + Data group 1 + Data group 4 Pattern 8: Data group 0 + Data group 2 + Data group 3 Pattern 9: Data group 0 + Data group 2 + Data group 4 Pattern 10: Data group 0 + Data group 3 + Data group 4 Pattern 11: Data group 0 + Data group 1 + Data group 2 + Data group 3 Pattern 12: Data group 0 + Data group 1 + Data group 2 + Data group 4 Pattern 13: Data group 0 + Data group 1 + Data group 3 + Data group 4 Pattern 14: Data group 0 + Data group 2 + Data group 3 + Data group 4 Pattern 15: Data group 0 + Data group 1 + Data group 2 + Data group 3 + Data group 4

[0047] 7, a second summary table 102 is created, which is a summary table corresponding to each of the data group patterns 1 to 15. The second summary table 102 includes items of each data group forming the above patterns and quality results of each development project. The second summary table 102 corresponds to the first quality control data, which is learning data input to the creation unit 11.

[0048] (Creating candidate predictive models) Next, as shown in Figure 8, the first quality control data (second summary table 102) created as described above is input into the creation unit 11 to create multiple prediction model candidates 1 to 15 corresponding to each data group pattern 1 to 15 (ST101).

[0049] The created prediction model candidates 1 to 15 are stored in a database 20 and are read out as needed to execute the various processes described below.

[0050] (Judgment Department) The determination unit 12 determines the prediction accuracy of the plurality of prediction model candidates 1 to 15 using the second quality control data, the evaluation results of which are known, as input information.

[0051] The first quality control data is used as the second quality control data. This can improve the prediction accuracy of each prediction model candidate. However, this is not limited to this, and data other than the first quality control data may be used as the second quality control data to improve the appropriateness of the prediction accuracy.

[0052] Fig. 9 is an explanatory diagram of the processing method in the determination unit 12. As shown in Fig. 9, the evaluation items (items 1 to 30) of each project in the first summary table 101 are used for the second quality control data. The determination unit 12 uses the prediction model candidate 1 as an evaluation tool to evaluate the quality prediction of project A based on each evaluation item of project A on a two-level scale of "good" and "bad." The example in Fig. 9 shows that the quality prediction result is "good."

[0053] Similarly, the determination unit 12 performs quality prediction for other projects B to Z using the prediction model candidate 1 as an evaluation tool. The results are shown in FIG. 10. FIG. 10 shows a comparison of the predicted results determined for each of projects A to Z with the actual quality results. The determination unit 12 determines a "match" when the quality result and the predicted quality result are both "◯" or both "×", and determines a mismatch otherwise. The determination unit 12 then calculates the percentage of the determination results that are "match" for all projects A to Z, and calculates this calculated value as the prediction accuracy. For example, when the total number of projects is 100 and the number of "match" determination results is 55 (the number of "mismatch" determination results is 45), the prediction accuracy is 55%.

[0054] The determination unit 12 performs the above determination process to predict the quality of all projects A to Z using the other prediction model candidates 2 to 15, and calculates the prediction accuracy of each of these prediction model candidates. The results are shown in FIG. 11. In this way, the determination unit 12 determines the degree of agreement between the prediction result obtained using the first quality control data as input information and the final quality result included in the first quality control data for each software development project. The determination results by the determination unit 12 may be displayed on the input / output unit 30 (display unit).

[0055] (Extraction part) The extraction unit 13 extracts at least one prediction model candidate from the plurality of prediction model candidates 1 to 15 as a representative prediction model based on the judgment results of the prediction accuracy of the plurality of prediction model candidates 1 to 15 (ST103).

[0056] In this embodiment, three candidate prediction models, for example, 5, 9, and 14, are extracted as representative prediction models from the candidate prediction models 1 to 15 in order of highest prediction accuracy (FIG. 11). For example, if the three extracted representative prediction models were created based on past deliverables up to the detailed design process, they are stored in database 20 as prediction models for quality evaluation of deliverables (process quality data) of all processes up to the detailed design process in the ongoing software development project.

[0057] Although only one representative prediction model may be extracted, in this embodiment, three representative prediction models are extracted. By using these multiple representative prediction models to evaluate the quality of the process quality data to be evaluated, the results of evaluation from various perspectives can be reflected in the evaluation results, compared to when only one representative prediction model is used, thereby improving the accuracy of the evaluation of the quality of the process quality data.

[0058] For example, if the probability that the predictions of three representative prediction models will be correct is 70% each (a 30% probability of incorrect prediction), the probability that all of the quality evaluations predicted by the three representative prediction models will be incorrect is 2.7% (30% x 30% x 30%). Therefore, by increasing the number of representative prediction models and performing quality predictions, even if the prediction of one representative prediction model is inappropriate, the prediction results of the other prediction models can be used as a reference, thereby increasing the probability of performing a more appropriate quality evaluation than when making a prediction using a single representative prediction model. Note that the number of representative prediction models is not limited to three, and may be two, four, or more.

[0059] Furthermore, even if the evaluation result of the process quality data when one representative prediction model A is used is "Good," there may be an "X" in one of the evaluation results of the process quality data when multiple representative prediction models are used. In this case, if the evaluation result is "Good" when a single representative prediction model is used, the process quality data may be evaluated as having no problems, but if the evaluation result of at least one of the multiple representative prediction models includes an "X," it can be evaluated that there is some kind of problem with the process quality data, allowing for a more objective final evaluation of the process quality data.

[0060] Furthermore, the method for extracting a representative prediction model may take into consideration not only the high prediction accuracy described above, but also the combination of data groups (the above-mentioned data groups 0 to 4) that form the representative prediction model. For example, consider a case where three prediction model candidates are extracted as representative prediction models from four prediction model candidates A to D. If prediction model candidate A is formed from three data groups 0, 1, and 2, prediction model candidate B is formed from two data groups 0 and 3, prediction model candidate C is formed from three data groups 0, 2, and 3, and prediction model candidate D is formed from three data groups 0, 3, and 4, prediction model candidate C overlaps with prediction model candidates A and B for data groups 2 and 3. Therefore, selecting the combination of prediction model candidates A, B, and D as the three representative prediction models to be extracted rather than the combination of prediction model candidates A, B, and C will provide a higher comprehensiveness of evaluation items.

[0061] By repeatedly executing the above process (ST101 to 103) for each development process, three representative prediction models are extracted for each development process and stored in the database 20. This allows prediction models for quality evaluation for each development process to be obtained.

[0062] (Evaluation Department) The evaluation unit 14 evaluates the quality of the third quality control data to be evaluated using the three representative prediction models extracted by the extraction unit 13 (ST104).

[0063] The third quality control data is process quality data, which is the deliverables of each development process in a software development project currently underway (hereinafter also referred to as project α). This process quality data has evaluation items similar to those of the first summary table 101 described with reference to Fig. 3, for example, and includes quality data (quality data for each development process) created for each of these evaluation items.

[0064] The evaluation unit 14 uses the multiple representative prediction models extracted by the extraction unit 13 to perform quality prediction (simulation) of process quality data (third quality control data), which is the deliverable of each development process in project α, as shown in Fig. 12, and outputs the evaluation result in two stages, for example, "o" or "x". The evaluation result by the evaluation unit 14 is displayed on the input / output unit 30 (display unit) and also stored in the database 20.

[0065] For example, if the above process quality data is a deliverable from the detailed design process in project α, the evaluation unit 14 reads out three representative prediction models for the detailed design process from the database 20, performs quality prediction of the process quality data using these representative prediction models, and outputs the evaluation results for each representative prediction model (three "◯"s in the example of Figure 12).

[0066] Similarly, for the process quality data of the unit test process and the integration test process, the evaluation unit 14 reads out three corresponding representative prediction models for the unit test process and the integration test process from the database 20, performs quality prediction of the process quality data using these representative prediction models, and outputs the evaluation results for each representative prediction model (in the example of Figure 12, one ``O'' for the unit test process and two ``O'' for the integration test process).

[0067] In this embodiment, the final quality evaluation of the process quality data to be evaluated is performed by an operator. The operator performs the final quality evaluation of the process quality data based on the judgment results of each representative prediction model in each development process. For example, if the prediction results are more "good" than "bad," or if all the prediction results are "good," the process quality data can be evaluated as a good product.

[0068] In order to improve the accuracy of predictions, simulations may be performed on past development projects before the actual predictions are made, and the trends in the evaluation results may be understood. For example, the types of trends when there are three representative prediction models are as follows: "Good" predictions have a higher accuracy rate than "bad" predictions The results of all three models are likely to be biased towards either "Yes" or "No." If all three models are "OK" in each development process, the accuracy rate is 80%. If all three models are "X" in the integration test process, the accuracy rate is 80%. Based on the above-mentioned types of tendencies, the worker can comprehensively judge the validity of the process quality data to be evaluated.

[0069] (summary) As described above, according to this embodiment, a representative prediction model is extracted from among candidate prediction models created using multiple quality control data whose evaluation results are known, and the representative prediction model is used to predict the quality of the quality control data to be evaluated, thereby increasing the reliability of the prediction results.

[0070] This makes it possible to more accurately evaluate the validity of each development process up to the development process being evaluated (design validity in the design process, program validity in the programming process, and test validity in the testing process) during the design stage, thereby significantly reducing the frequency of defects occurring in the final process and enabling improved work efficiency and shorter delivery times in software development.

[0071] In addition, since process quality data (first quality control data) related to two or more software development projects is used as the learning model, the versatility of the evaluation results is increased, thereby further improving the accuracy of the quality evaluation.

[0072] Furthermore, by creating multiple candidate prediction models corresponding to multiple types of data group patterns that are created by arbitrarily combining multiple data groups according to the development process, the amount of data required for training each candidate prediction model can be reduced, and the time required to create the candidate prediction models can be shortened. In particular, because the training data for each candidate prediction model created includes at least the required data groups in common, significant variations in prediction accuracy can be suppressed.

[0073] Although the embodiments of the present invention have been described above, it goes without saying that the present invention is not limited to the above-described embodiments and that various modifications can be made.

[0074] For example, in the above embodiment, software development for a fire department radio system was used as an example, but the type is not particularly limited, and the present invention can be applied to software development for various systems such as fire department radio systems, disaster prevention radio systems, and distribution services.

[0075] In addition, in the above embodiment, the first quality control data for creating a candidate prediction model was used as the second quality control data input during the judgment process in the judgment unit 12, but this is not limited to this, and quality control data of other software other than the quality control data used as training data may also be used.

[0076] Furthermore, in the above embodiment, the evaluation result of the evaluation object output by the evaluation unit 14 is in two levels, "◯" and "×", but this is not limited to this, and the evaluation result may be in three or more levels, or the evaluation result may be output as a more specific numerical value. [Explanation of symbols]

[0077] 1. Information processing system 10...Information processing device 11...Creation Department 12…Judgment section 13...Extraction part 14...Evaluation Department 20...Database 30…Input / output section

Claims

1. a creation unit that creates a plurality of prediction model candidates using the first quality control data, the evaluation results of which are known, as learning data; a determination unit that determines the prediction accuracy of the plurality of prediction model candidates using second quality control data, the evaluation results of which are known, as input information; an extracting unit that extracts at least one prediction model candidate from the plurality of prediction model candidates as a representative prediction model based on a result of determining the prediction accuracy of the plurality of prediction model candidates; an evaluation unit that evaluates the quality of third quality control data to be evaluated using the representative prediction model; An information processing device comprising:

2. 2. The information processing device according to claim 1, The first quality control data includes quality data and quality results including aggregated values ​​regarding defects aggregated for each evaluation item in each development process of two or more software development projects. Information processing device.

3. 3. The information processing device according to claim 2, Each evaluation item of the first quality control data is classified into a plurality of data groups each consisting of a combination of similar evaluation items; The creation unit creates a plurality of prediction model candidates corresponding to a plurality of types of data group patterns obtained by arbitrarily combining the plurality of data groups according to the development process. Information processing device.

4. 3. The information processing device according to claim 2, The first quality control data includes both quality data in which the quality result is good and quality data in which the quality result is bad. Information processing device.

5. 4. The information processing device according to claim 3, The plurality of types of data group patterns include an essential data group that is a data group common to each data group pattern. Information processing device.

6. 4. The information processing device according to claim 3, The extraction unit extracts three or more prediction model candidates from the plurality of prediction model candidates as the representative prediction model. Information processing device.

7. 3. The information processing device according to claim 2, the second quality control data is the first quality control data, The determination unit determines a degree of agreement between a prediction result obtained using the first quality control data as input information and the quality result included in the first quality control data for each of the two or more software development projects. Information processing device.

8. 8. The information processing device according to claim 7, The extraction unit extracts, as the representative prediction model, a prediction model candidate whose degree of agreement of the prediction results in all software development projects is equal to or greater than a predetermined value. Information processing device.

9. A quality evaluation method executed in an information processing device, comprising: creating a plurality of candidate prediction models based on first quality control data whose evaluation results are known; determining the prediction accuracy of the plurality of candidate prediction models using second quality control data, the evaluation results of which are known, as input information; extracting at least one candidate prediction model as a representative prediction model from the candidate prediction models based on the judgment results of the prediction accuracy of the candidate prediction models; The representative prediction model is used to evaluate the quality of the third quality control data to be evaluated. Quality assessment methods.

Images