JP2011198300A - Process improvement measure evaluation device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a process improvement measure evaluation device and a method where a return caused by a fault and a cost to be required for improvement measures are incorporated into an evaluation target.SOLUTION: The process improvement measure evaluation device includes: a process flow model 12 for generating a model where man-hour in the case without a fault, the number of faults to be modified in the process, and a man-hour required per one fault in the process are regulated concerning each process of a project; a setting part 14 for setting in the model, the number of the faults and the man-hour concerning each process based on a teacher project with the past results recorded therein; a measures influence estimating part 17 for changing the number of the faults in each process which is set in the model based on the process improvement measure information with a project plan value changed therein, and updating the man-hour of each process after the process improvement measures based on the changed number of the faults; and an output part 16 for outputting a process improvement measure evaluation value including the updated information.

Description

  The present invention relates to a process improvement measure evaluation apparatus and method.

  When developing a software system as a project, it is necessary to go through a series of processes such as system design, software production, and operation test. On the other hand, if a problem occurs in each process under development, rework occurs. The rework means going back to the previous process according to the cause of the defect and executing the process again from there.

  So far, various methods have been proposed for evaluating the project by collecting development results such as man-hours required for each process of the development project in order to devise an improvement plan for efficiently operating the project.

  In the invention disclosed in Patent Document 1, not only the number of defects in a project process can be evaluated, but also the degree of outflow to a downstream process can be evaluated. As a result, the detection capability in a design process and an early test process can be improved. A method that can evaluate the degree of improvement is proposed.

  The invention described in Patent Document 2 discloses a method for evaluating the influence of a modification of a plan in each process on a project by holding the influence related to the change of the process in a scenario format.

JP 2007-316728 A JP 2005-032079 A

  However, in the technique described in Patent Document 1, since the influence of rework is not taken into consideration in the process, when the process is affected by the occurrence of many rework, the accuracy of the project evaluation is lowered.

  In the evaluation method disclosed in Patent Document 2, attention is paid only to the number of defects. Therefore, even when the same evaluation is made, there are cases where the cost (for example, man-hour) required for the process improvement measures to be implemented greatly increases and when it does not. Therefore, it is not sufficient to evaluate only the number of defects.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide a process improvement measure evaluation apparatus and method in which rework due to a defect and costs required for an improvement measure are incorporated in an evaluation target.

  In the present invention for solving the above problems, the number of man-hours when there is no defect, the number of defects to be corrected in this process, and the man-hour required for each defect in this process are specified for each process of the project. A process flow model part that creates a model, a setting part that sets the number of defects and man-hours for each process from the teacher project that has been recorded in the past, and a process improvement measure that changes the planned value of the project The policy impact estimation unit that changes the number of defects in each process set in the model based on information, and updates the man-hours of each process after a process improvement measure based on the changed number of defects The process improvement measure evaluation apparatus includes an output unit that outputs a process improvement measure evaluation value including information.

  In addition, the present invention creates a model in which the man-hour when there is no defect for each process of the project, the number of defects to be corrected in this process, and the man-hour required per defect in this process, The defect of each process set in the model based on the process improvement measure information in which the number of defects and man-hours for each process are set in the model from the teacher project whose results have been recorded in the past and the planned value of the project is changed This is a process improvement measure evaluation method that changes the number of cases, updates the man-hours of the respective steps after the process improvement measure based on the changed number of defects, and outputs a process improvement measure evaluation value including the updated information.

  According to this invention, it is possible to evaluate the process improvement measure by incorporating the rework due to the defect and the cost required for the improvement measure into the evaluation target.

The figure which shows the structure of a process improvement measure evaluation system. The figure which shows the content of man-hour DB. The figure which shows the content of defect DB. The figure which shows the content of project DB. The figure which shows a process flow model. The figure which shows the data correction classification model different from a manual return classification model. The flowchart which shows the general | schematic operation | movement procedure of a malfunction information extraction part. The figure which shows a malfunction matrix. The flowchart which shows the operation | movement procedure of the outline of a man-hour information extraction part. The figure which shows the man-hour data of a project. The figure which shows the man-hour data of the project in which only the total man-hour is registered. The figure which shows the man-hour data of the project in which the project where the initial man-hour and the return man-hour are registered and the project where only the total man-hour is registered are mixed. The flowchart which shows the procedure which calculates a return man-hour and an initial man-hour from total man-hour. The figure which illustrates the produced defect matrix. The figure which extracted the malfunction of the comprehensive design and the mixed process in the discovery process. The figure which showed the subtraction result of a discovery date and a confirmation date. The flowchart which shows the procedure at the time of registering in matrix form. The figure which shows a matrix format. The figure which shows a matrix format. The figure which shows a matrix format. The figure which shows a matrix format. The figure which shows a matrix format. The figure which shows a matrix format. The figure which shows the content of review test data. The flowchart which shows the process sequence of a measure influence estimation part. The figure for demonstrating the regression analysis of a coverage rate and a detection rate. The flowchart which shows the operation | movement of the outline of a process improvement measure evaluation apparatus.

[First Embodiment]
FIG. 1 is a diagram showing a configuration of a process improvement measure evaluation system. The process improvement measure evaluation system includes the process improvement measure evaluation device 1 and the input / output device 2 according to the first embodiment of the present invention.

The process improvement measure evaluation apparatus 1 includes a measure information input unit 10, a teacher project selection unit 11, a process flow model unit 12, a defect information extraction unit 13, a manhour information extraction unit 14, a manhour calculation unit 15, and a result as a unit that executes processing. An output unit 16, a measure influence estimation unit 17, and a control unit 20 are provided.
Moreover, the process improvement measure evaluation apparatus 1 has a man-hour DB 21, a defect DB 22, a project DB 23, and a review test DB 24 as databases (DB) for storing data.
The input / output device 2 inputs a user instruction to the process improvement measure evaluation device 1 and displays information output from the process improvement measure evaluation device 1.

The measure information input unit 10 inputs information necessary for performing the evaluation. The teacher project selection unit 11 selects a project similar to the target project as a base when performing process improvement measure evaluation. The process flow model unit 12 models the transition of the process and the manner of return when a failure occurs as a process flow model.
The defect information extraction unit 13 extracts parameters related to defects in the process flow model based on the data related to defects. The man-hour information extracting unit 14 extracts parameters relating to the man-hours of the process flow model. The man-hour calculation unit 15 calculates the total man-hour required for the project. The result output unit 16 outputs the evaluation result to the monitor of the input / output device 2. The measure impact estimation unit 17 calculates an evaluation value representing the result when the process improvement measure in which the project plan value is changed is adopted. The control unit 20 controls the operation of each unit of the process improvement measure evaluation device 1 in an integrated manner.

  Next, as a premise for explaining the operation of the process improvement measure evaluation apparatus 1, terms and data contents used in the process improvement measure evaluation apparatus 1 will be described with reference to the configuration of each DB.

  FIG. 2 is a diagram showing the contents of the man-hour DB 21. The man-hour data stored in the man-hour DB 21 is data in which project ID, data man-hour, logic man-hour, total man-hour, initial man-hour, and rework man-hour in each development process are recorded for each project. This man-hour data can be searched using the project ID as a keyword.

  The project ID is data for identifying a project, and different projects have different project IDs. Data man-hours and logic man-hours are values held only by a software development project that handles a large amount of data in the project.

For example, a system targeting station equipment handles a large amount of information (data) such as stations and routes. The total man-hour required to develop these data is the data man-hour. In addition, the total man-hour required for developing the system operation is the logic man-hour.
In the case of a ticketing machine, the total man-hours required for designing, mounting, and verifying display information to be printed on a ticket and information to be added are data man-hours. Further, the total man-hours required for designing / implementing / verifying system operation / response for user operations such as button presses and coin insertions are the logic man-hours. The data man-hour of the i-th project IDi is represented as Di, and the logic man-hour is represented as Li.

The total man-hour in each development process is the total man-hour required in that process. The rework man-hour is a man-hour required in the process to correct one defect when a defect occurs. The initial man-hour is the man-hour required first for the process, and is obtained by subtracting the total man-hour required for reworking from the total man-hour.
The total man-hours, rework man-hours, and initial man-hours for the process p of the project IDi are Sip, Tip, and Iip, respectively. When the number of defects returned to the process p of the project IDi is Bip, Sip = Tip × Bip + Iip is established.

  FIG. 3 is a diagram showing the contents of the defect DB 22. The defect data stored in the defect DB 22 includes, for each project, a project ID, a defect ID, a mixing process, a discovery process, a discovery date, a confirmation date, and data correction identification information. This defect data can be searched by project ID.

  The defect ID is data for identifying a defect in the same project. Therefore, the same identification data may be used for different projects. The mixing process is a process in which a defect is mixed, and the discovery process is a process in which a defect is found. The discovery date is the date when the defect is found, and the confirmation date is the date when the defect is corrected and it is confirmed that the correction has been made correctly. In addition, the data correction identification information is information for identifying whether or not a defect only requires data correction in a project that handles a large amount of data.

  In the trouble ID 1 of the project ID 1 shown in FIG. 3, the mixing process is a basic design, the discovery process is a unit test, the discovery date is 2008/11/4, and the confirmation date is 2008/11/10. In the case of the defect ID1 of the project ID1, the defect can be corrected only by data correction.

  FIG. 4 is a diagram showing the contents of the project DB 23. The project data stored in the project DB 23 includes information indicating a project ID and a project type. As information representing the type of project, FIG. 4 uses a development rank in which projects are ranked by development scale. In addition, even if the information indicating the project type is not the development rank, any information indicating the project type discretely can be substituted.

Next, the process flow model will be described.
The process flow model is a model of a process transition state and a return state when a failure occurs. In this embodiment, a process flow model of a project based on the processes of system design, detailed design, basic design, manufacturing (mounting), unit test, combination test, and comprehensive test is considered.

  In the process flow model shown in FIG. 5, each process is treated as a separate process at the initial time and at the return time even if it is the same process. A white block in FIG. 5 represents an initial process of each process, and a hatched block represents a return process of each process. Further, Ti described in the white block means the initial man-hour in the process i, ti in the hatch block means the reworking man-hour in the process i, and bi means the number of defects reaching the reworking process in the process i. The solid line connecting the blocks represents the transition of the process, and the dotted line connecting the blocks represents the return due to the malfunction.

  Moreover, the white block described as "following process" represents the customer test and the process after release. In other words, this process flow model also takes into account rework of defects found after customer testing and release. Hereinafter, this process flow model is referred to as a return sorting model.

  FIG. 6 is a diagram showing a data correction classification model different from the above-described manual return classification model. In the data correction classification model, in software development that handles a large amount of data, defects that only require data correction are transferred to the data correction process without returning to the return flow. The data correction step is a step of correcting a defect related to the data value, and is a step of correcting only the change of the data value. Unlike other problems, data correction problems can be corrected in a very short time. Therefore, by distinguishing from other problems, the accuracy of the model can be improved.

  Ti described in the white block and ti described in the hatch block are the initial man-hours and the rework man-hours of the process i as in the rework classification model. bi (1 ≦ i ≦ 7) is the number of defects that pass through the return process of the process i that is not a data correction. In the hatch block described as “data correction”, td is the man-hour required for data correction, and bd is the number of defects in data correction. The value of td is set to a constant value based on past project results.

Next, the operation of the process improvement measure evaluation apparatus 1 according to this embodiment will be described.
FIG. 27 is a flowchart showing a schematic operation of the process improvement measure evaluation apparatus 1.

In step S <b> 501, the user (user) inputs teacher project calculation information using the measure information input unit 10.
The teacher project is a project that can be used as the base data for the evaluation of the process improvement measure among the projects whose results have been recorded in the past. Information necessary for teacher project calculation input by the user includes a project ID, a development rank shown in FIG. 4, and the like, and a data / logic ratio is a software development project that handles a large amount of data. The user needs to input at least one piece of information.

  Here, the data / logic ratio is information on the ratio of the amount of data development to the amount of logic development in a software development project. If you have experience in software development for the same domain, It is possible to estimate the ratio.

  In the present embodiment, when a similar past project is known as the teacher project calculation information, the user inputs a project ID. A plurality of project IDs can be input. On the other hand, when a similar past project is not known, the user inputs a development rank and a data / logic ratio representing the characteristics of the project as teacher project calculation information.

In step S502, the measure information input unit 10 checks whether the input information is a project ID. If Yes in step S502, that is, if the input information is a project ID, the process proceeds to step S504.
If No in step S502, that is, if the input information is not a project ID, the measure information input unit 10 outputs teacher project calculation information to the teacher project selection unit 11.

  In step S503, if the development rank and the data / logic ratio are input as the teacher project calculation information, the teacher project selection unit 11 searches the project DB 23 and the man-hour DB 21 to obtain the man-hour data of the project having the same development rank. Extract. Then, the ratio of the data man-hour and the logic man-hour is obtained, a project having a ratio close to the input data-logic ratio is selected as a teacher project, and the project ID is returned to the measure information input unit 10. If there are a plurality of projects having approximate ratios, all the project IDs are returned to the measure information input unit 10.

On the other hand, when only the development rank is input, the teacher project selection unit 11 returns all project IDs of the same development rank to the measure information input unit 10.
If only the data / logic ratio is input, the man-hour data of all projects is referred to and the ratio of the data man-hour and the logic man-hour is obtained. Then, a project close to the input data / logic ratio is selected as a teacher project, and the project ID is returned to the measure information input unit 10. If there are a plurality of projects having approximate ratios, all the project IDs are returned to the measure information input unit 10.

  When the project ID is obtained, the measure information input unit 10 outputs the project ID to the defect information extraction unit 13. In step S504, the defect information extraction unit 13 searches the defect DB 22, and refers to the defect data (FIG. 3) for the input project ID to calculate the return defect bi of the process flow model. Then, the value bi is set in the process flow model.

  The operation of the defect information extraction unit 13 will be described in detail.

FIG. 7 is a flowchart showing a schematic operation procedure of the defect information extraction unit 13.
In step S001, the defect information extraction unit 13 receives the project ID from the measure information input unit 10. In step S002, the defect information extraction unit 13 searches the defect DB 22 for reference to the received defect data of each project, and creates a defect matrix in step S003.

  FIG. 8 is a diagram showing a defect matrix. The defect matrix is a matrix in which a process in which a defect is found on the vertical axis, a process in which a defect is mixed on the horizontal axis, and the number of defects returned from the discovery process to the mixing process are written in each column.

  For example, b31 is the total number of defects found in the review performed in the detailed design process and returned to the system design process. b62 is the total number of defects found in the integration test process and returned to the basic design process. Since there is no rework to return to the subsequent process, b12, b13, b14, b23, b24, and b34 are always 0 in this defect matrix. In addition, since no defects are mixed in the test process, the mixing process on the horizontal axis extends to the mounting process.

  A method for creating this defect matrix will be described. At the time of step S001 described above, the value of each column of the defect matrix is set to zero. When the process flow model is a data correction classification model, the value of the number of return bd in the correction process is also set to zero. In the creation process in step S003 described above, in the case of the data correction classification model, the data correction identification information in the defect DB 22 is referred to, and 1 is added to bd if the data correction only requires the data correction. If it is not a defect that requires only data correction, the discovery process and the mixing process are referred to and 1 is added to the value of the corresponding defect matrix column. For example, if the mixing process is a detailed design and the discovery process is a unit test, the value b53 of the defect matrix in FIG.

  After the creation of the defect matrix for one project, it is checked in step S004 whether there is defect data for an unreferenced project. In the case of Yes in step S004, that is, when there is unreferenced defect data, the above-described processing of steps S002 to S003 is executed again. In the case of No in step S004, that is, when there is no unreferenced defect data, in step S005, each value of the defect matrix is divided by the number of projects to obtain an average value per project. When the process flow model is a data correction classification model, the value of bd is also an average value.

  In step S006, from the defect matrix obtained in step S005, the value of the number of defects bi (1 ≦ i ≦ 7) reaching the return process of step i required by the process flow model is calculated.

  The problem of returning to the system design rework process is that the mixing process is a system design problem. Therefore, b1 = Σ (1 ≦ j ≦ 8) bj1. The failure to return to the basic design rework process becomes a rework failure with the basic design between the mixing process and the discovery process. That is, b2 is a value obtained by adding the troubles mixed in the basic design, excluding the troubles found in the system design, from the number b1 of troubles returning to the system design reworking process which is the previous process. Therefore, b2 = b1−b11 + Σ (1 ≦ j ≦ 8) bj2. Similarly, b3 = b2−b21−b22 + Σ (1 ≦ j ≦ 8) bj3 and b4 = b3−b31−b32−b33 + Σ (1 ≦ j ≦ 8) bj4.

  Since b5 and subsequent steps are test steps and are not mixed steps, the number of defects found in the immediately preceding process is calculated from the number of defects returned from the immediately preceding return process. Therefore, b5 = b4-Σ (1 ≦ j ≦ 4) b4j, b6 = b5-Σ (1 ≦ j ≦ 4) b5j, b7 = b6-Σ (1 ≦ j ≦ 4) b6j. Further, the number of defects remaining in the subsequent processes can be calculated by Σ (1 ≦ j ≦ 4) b8j.

  Returning to FIG. 27, in step S505, the man-hour information extracting unit 14 searches the man-hour DB 21 to refer to the man-hour data (FIG. 2) of the project ID. Then, using this man-hour data, the initial man-hour Ti and the rework man-hour ti of the process flow model are calculated, and the calculated values are set in the process flow model.

  The operation of the manhour information extracting unit 14 will be described in detail.

FIG. 9 is a flowchart showing a schematic operation procedure of the man-hour information extracting unit 14.
In step S <b> 101, the man-hour information extraction unit 14 receives one or more project IDs from the measure information input unit 10. In step S102, the man-hour data of each project ID received by searching the man-hour DB 21 is referred to. In step S103, the man-hour information extracting unit 14 checks whether or not the man-hour data includes registration of rework man-hours and initial man-hours in each process.

  In the case of Yes in step S103, that is, if there is registration of the return man-hour and the initial man-hour, the registered value is obtained in step S104. In the case of No in step S103, that is, if there is no registration, the return man-hour and the initial man-hour are calculated from the total man-hour in step S105. The procedure of step S105 will be described later in detail.

  In step S106, it is checked whether or not the return man-hours and initial man-hours have been acquired for all the received project IDs. In the case of No in step S106, that is, if there is a project for which the rework man-hour and the initial man-hour are not acquired in the received project ID, the process returns to step S102 and the man-hour data is referred again. In the case of Yes in step S106, that is, when the return man-hours and initial man-hours of each process of all projects are acquired, in step S107, the average of the acquired initial man-hours and rework man-hours is obtained for the project. The value is substituted into the process flow model parameters Ti and ti of the corresponding process.

This process will be described using a specific example.
For example, it is assumed that the received project IDs are a and b, and the manhour data of those projects are the values shown in FIG. In this case, the procedure of the flowchart of FIG. 9 is Yes in step S103 and Yes in step S106. As a result, the man-hour information extraction unit 14 sets Ti = (Iai + Ibi) / 2 and ti = (Tai + Tbi) / 2 in the process flow model.

On the other hand, as shown in FIG. 11, when only the total man-hour is registered in the man-hour data, the procedure of the flowchart of FIG. 9 is as follows: No in step S103 twice and Ti and ti in the process of step S105. Calculate the value.
In addition, as shown in FIG. 12, when a project in which the initial man-hours and rework man-hours are registered and a project in which only the total man-hours are registered are mixed, the registered ones are used for the registration. Those that are not calculated are calculated in step S105.

Subsequently, the process of step S105 will be described in detail. The calculation of the return man-hour and the initial man-hour from the total man-hour, which is the process of step S105, is performed using the defect data in the defect DB 22.
FIG. 13 is a flowchart showing a procedure for calculating the return man-hour and the initial man-hour from the total man-hour.

  In step S201, the defect data is referenced from the project ID, and a defect matrix for the project is created. Since the defect matrix creation method has already been described, detailed description thereof will be omitted. FIG. 14 is a diagram illustrating the created defect matrix. In step S202, the column that holds the largest value is selected from the columns of the created defect matrix. In FIG. 14, since the discovery process has the largest value (10) in the comprehensive test column and the mixing process has the largest value (10), this column is selected. If there are multiple columns giving maximum values, select one using random numbers.

  Subsequently, in step S203, an equation of the process flow model ti is created using the confirmation date and the discovery date of the defect data corresponding to the selected field. FIG. 15 is a diagram in which, from the defect data, the discovery process in the project ID 10 is a comprehensive test, and the mixing process is a detailed design defect. According to the process flow model, the number of man-hours required for returning from the comprehensive test process to the detailed design process is t3 + t4 + t5 + t6. Here, t7 is not included in the man-hours required for reworking because t7 includes a value other than this cause. On the other hand, in each defect shown in FIG. 15, the value obtained by subtracting the discovery date from the confirmation date is the time required for the defect to be corrected, and can be regarded as the man-hours required for each defect to return.

  FIG. 16 is a diagram showing the subtraction result of the discovery date and the confirmation date based on FIG. In this example, it is assumed that the defect corrected on that day, such as defect ID 4, took one day to return. This can be regarded as 0. Moreover, the holiday (2008/11/8, 11/9) is not counted like defect ID7. In this way, even when returning from the same comprehensive test to detailed design, the correction time varies depending on the defect. An average value is adopted as the reworking man-hour ti of the process flow model in order to reduce this variation. In the example shown in FIG. 16, an average value = 7.8 is obtained. That is, an equation of t3 + t4 + t5 + t6 = 7.8 can be obtained. However, in this equation, since the right side is the number of days and the left side is the man-hour (person × time), the unit is adjusted by applying an appropriate coefficient. In the following description, it is assumed that the coefficient is 1.

  In the calculation of the average value, outliers may be removed. In the example shown in FIG. 16, the discovery date-confirmation date of the defect ID 17 is a very large value compared to other defects. Whether the value is off can be detected using the standard deviation. For example, a value that is three times or more away from the average than the standard deviation can be defined as an outlier. In the case of FIG. 16, since the average is 7.8 and the standard deviation is 8.71, the defect ID 17 can be regarded as an outlier.

  However, the defect detected as described above is not always an outlier with low reliability. It is also possible that it took time for a bug that was difficult to fix. Therefore, by registering the importance of the defect in the defect data, a method of removing the data having a low importance from the data whose detection date-confirmation date is not in the direction of increasing is regarded as having low reliability. be able to.

  As described above, the ti equation of the process flow model is created using the confirmation date and the discovery date of the defect data. The equations obtained in this way are managed in the form of matrix operation as shown in FIG.

  Subsequently, in step S204, it is checked whether all the reworking man-hours ti are obtained from the obtained equation. According to the flowchart of FIG. 13, since there is only one expression in the matrix of FIG. 18, it is not possible to obtain all the reworking man-hours ti (No in S204). In step S205, if there is one or more unselected columns in the defect matrix (FIG. 14) (Yes in S205), the processing from step S202 is repeated and executed again.

  A plurality of equations are created by the above procedure and managed in the form of a matrix. FIG. 17 is a flowchart showing a procedure when registering in the matrix format.

  In step S301, it is checked whether there is a case in which the left side of the row to be registered is included in the left side of the already registered row. If Yes in step S301, that is, if there is a case where the left side of the line to be registered is included in the left side of the already registered line, in step S307, an attempt is made to register from the included line at the position of the included line. Subtract lines that are present. Next, in step S308, it is confirmed whether the right side is negative. In the case of Yes in step S308, that is, in the case of being negative, since the reworking man-hour ti is a non-negative number, the registration of the equation is stopped in step S309. Here, when canceling registration, the matrix is returned to the state before the subtraction without reflecting the already subtracted result. If No in step S308, that is, if the right side is not negative, the result of subtraction is reflected in the matrix and the process returns to step S301.

  If No in step S301, that is, if there is no case where the left side of the registered line is included in the left side of the already registered line, the left side of the already registered line is the left side of the registered line in step S302. To see if they are equal. If Yes in step S302, that is, if equal, the registration of the equation is stopped in step S309. Here, when canceling registration, the matrix is returned to the state before the subtraction without reflecting the already subtracted result.

  In the case of No in step S302, that is, if they are not equal, it is checked in step S303 whether there is a case where the left side of the already registered line is included in the left side of the registered line. If No in step S303, that is, if there is no such case, the equation is registered in step S310. If Yes in step S303, that is, if there is such a case, the line to be registered is subtracted from the already registered lines in step S304.

  In step S305, the sign of the right side of the subtraction result is checked. If Yes in step S305, that is, if the right side of the row to be registered becomes negative, the equation registration is canceled in step S309. Here, when canceling registration, the matrix is returned to the state before the subtraction without reflecting the already subtracted result. If No in step S305, that is, if the right side of the registered line is not negative, it is checked in step S306 whether the left side of the already registered line is equal to the left side of the registered line. If Yes in step S306, that is, if there is such a case, the process returns to step S304. If No in step S306, that is, if not, the process returns to step S301.

  Next, processing according to the above flow chart will be described using a specific matrix as an example.

  For example, when an equation represented as t4 + t5 + t6 = 8.5 is registered in the matrix format in the state shown in FIG. 18, it is included in the already registered rows. That is, since the first row of the matrix represents t3 + t4 + t5 + t6 = 7.8, the expression on the left side of the above equation is included. Therefore, subtraction is performed from the first row as shown in FIG. Then, the right side becomes a negative number. Therefore, the equation t4 + t5 + t6 = 8.5 is not registered. In this example, according to the flowchart of FIG. 17, “Yes” is determined in step S301, and “Yes” is determined in step S308. Therefore, registration is stopped in step S309.

  On the other hand, when an equation represented by t4 + t5 + t6 = 7.0 is registered in the matrix format shown in FIG. 18, the right side is positive as a result of subtraction from the first row. Therefore, after subtracting the first row, the equation is registered in the second row, which is the next row. The result is the line example shown in FIG. In this example, according to the flowchart of FIG. 17, “Yes” is determined in step S301, and “No” is determined in step S308. Therefore, the process returns to step S301. In this example, No is obtained in step S301, No is obtained in step S302, and registration is performed through No in step S303.

  Further, for example, when an equation expressed as t2 + t3 + t4 + t5 + t6 = 7.5 is registered in the matrix form shown in FIG. 18, the subtraction shown in FIG. 21 is performed with respect to the first row, and the right side becomes a negative value. . Therefore, registration of the equation t2 + t3 + t4 + t5 + t6 = 7.5 is stopped. In this example, according to the flowchart of FIG. 17, No in step S301, No in step S302, Yes in step S303, Yes in step S305, and registration is stopped.

  On the other hand, when the equation expressed as t1 + t2 + t3 + t4 + t5 + t6 = 9.0 is registered in the matrix format shown in FIG. 22, the subtraction shown in FIG. 23 is performed between the first row and the third row of the matrix. The second row has the same value. Therefore, registration is canceled. According to the flowchart of FIG. 17, No in step S301, No in step S302, Yes in step S303, No in step S305, No in step S306, No in step S301, and Yes in step S302. This is an example of canceling registration.

  Returning to FIG. 13, as described above, in step S204, it is managed whether or not the created equation can be solved. When the number of registered equations is equal to the number of ti, all values of ti can be obtained. Therefore, if Yes in step S204, Ti is calculated using ti obtained as the solution of the equation in step S207. In step S207, it is assumed that bi has already been obtained. The man-hour DB 21 is searched, and the total man-hour TSi in each development process is obtained from the project ID. Then, since the relationship of TSi = Ti + bi × ti is established, the value of Ti can be calculated from this equation.

  Returning to FIG. 27, the processing of step S504 and step S505 is completed as described above, but the order of executing step S504 and step S505 may be reversed.

  When the process flow model parameter setting is completed, the total man-hour is calculated by the man-hour calculating unit 15 and the calculation result is displayed by the result output unit 16 in step S506.

The operation of the man-hour calculation unit 15 will be described.
The man-hour calculation unit 15 calculates the total man-hour required for the project. The calculation method differs depending on the process flow model, but it is common to obtain the calculation from the total of the initial man-hours and the total of the rework man-hours. In the return sorting model, the total man-hour is obtained by calculating Σ1 ≦ i ≦ 7 (Ti + bi · ti). In the data correction classification model, the total man-hour is obtained by calculating Σ1 ≦ i ≦ 7 (Ti + bi · ti) + bd · td.

As a result, the operation of the output unit 16 will be described.
The result output unit 16 displays the calculation result on the monitor of the input / output device 2 and notifies the user. In the case of the above calculation result, the information to be notified is the total man-hours, Ti and ti values of the process flow model, defect matrix, the number of defects remaining in the subsequent processes, the total man-hours of each process, and the total man-hours of each process The total value of

The user refers to the output result and examines a project improvement measure.
Process improvement measures include, for example, changing the review test time and test process coverage (to be described later) to estimate changes in the Ti, ti, and bi values of the process flow model and Investigate. Thus, the process improvement measure evaluation apparatus 1 can be used to examine a project promotion method for reducing the total man-hours.

  Note that the time required for review in each design / mounting process, the test coverage rate in each test process, the number of test execution steps, the test design time, and the like are recorded in the review test DB 24 as review test data for each project. FIG. 24 is a diagram showing the contents of the review test data.

  The coverage rate applied as a process improvement measure is preferably different depending on the test process. Code coverage can be used for unit tests, and system coverage and combination coverage can be used for combination tests and comprehensive tests.

Code coverage is a measure of how much testing has been performed on the source code base. This standard includes C0 (instruction coverage) indicating the proportion of statements executed in the test among all statements, and C1 (branch coverage) indicating the proportion of branches executed in the test among all branches. and so on.
System coverage is a standard indicating how many tests have been executed based on the module to be tested. The standard includes S0 indicating the ratio of modules executed in the test among all modules, S1 indicating the ratio of modules executed in the test among all module call instructions, and the like.

  The combination coverage is a test standard that indicates a ratio of combinations verified by a test among all possible combinations. For example, when viewing the coverage of combinations between two functions, the coverage ratio is (number of combinations of two functions verified in a test) / (number of combinations between all two functions).

  In step S508 of FIG. 27, the user inputs process improvement measure information that is a planned value of the project from the measure information input unit 10. As process improvement measure information, the review time or coverage rate is handled below. In step S <b> 509, the measure information input unit 10 outputs process improvement measure information to the measure impact estimation unit 17.

  The measure impact estimation unit 17 quantitatively estimates the amount of change in the number of defects bi and Ti when the review time or the coverage rate is changed as process improvement measure information. In the estimation, the defect detection rate is used. The defect detection rate is calculated by (defect detected in step i) / (defect to be detected in step i) and is an index indicating how many defects to be detected in step i are detected.

The defect detection rate can be calculated from the defect matrix (FIG. 8). If it is a design / manufacturing process, a defect to be mixed should be found in the review of the same process. Therefore, the defect detection rate of process i = Σbii / {Σ (i <j ≦ 8) bji).
In the test process, manufacturing unit and detailed design should be detected for unit tests, basic design should be detected for combined tests, and system design failures should be detected for comprehensive tests. Therefore, the defect detection rate in each test is defined by the following equation. The defect detection rate of the unit test is (b53 + b54) / (b53 + b54 + b63 + b64 + b73 + b74 + b83 + b84). The failure detection rate of the binding test is b62 / (b62 + b72 + b82). The defect detection rate of the comprehensive test is b71 / (b71 + b81).

FIG. 25 is a flowchart showing the processing procedure of the measure impact estimation unit 17.
In step S401, it is checked whether the process improvement is performed for review or test.

  If the test is a target in step S401, regression analysis is performed between the coverage rate and the test design man-hour, and the coverage rate and the test execution man-hour in step S402. In step S403, the test design man-hour and the test design man-hour required to cover the input coverage rate are estimated using the regression analysis result. Next, in step S404, regression analysis is performed using the coverage rate and the defect detection rate. In step S405, the regression analysis result is used to estimate the number of defects detected when the test is performed with the input coverage rate.

  On the other hand, if the review is a target in step S401, regression analysis is performed using the review time and the defect detection rate in step S406. In step S407, the regression analysis result is used to estimate the number of defects detected when the input review time is applied.

  Regression analysis is a statistical technique that examines the relationship between a variable as a result and a variable as a factor, and estimates the relationship between variables. In the present embodiment, the review time and defect detection rate (step S406), coverage rate and test design man-hour (step S402), coverage rate and test execution man-hour (step S402), coverage rate and defect detection rate (step S404) Is estimated.

Subsequently, the relationship between the above-described variables used in the regression analysis will be described.
With regard to the relationship between the review time and the defect detection rate (step S406), many defects that are easy to find in a review are found within a short period of time, and defects that are difficult to find in a review tend to be found sporadically thereafter. Therefore, the defect detection rate does not decrease with respect to the increase in the review time, but the increase rate decreases.

With regard to the relationship between the coverage rate and the test design man-hour (step S402), it is easy to perform test design so that the coverage rate is increased with a low coverage rate, but the coverage rate is further increased with a high coverage rate. Tend to be difficult. Therefore, as the test design man-hour increases, the coverage rate does not decrease, but the increase rate decreases.
Similarly, the relationship between the coverage rate and the test execution man-hour (step S402) is similar to the increase in test execution man-hours, but the coverage rate does not decrease but the increase rate decreases. On the other hand, regarding the relationship between the coverage rate and the defect detection rate (step S404), if there is no characteristic that is unevenly distributed in the presence of defects, it is considered that the defect detection rate increases proportionally as the coverage rate increases.

  Thus, linear regression can be applied to the regression analysis between the coverage rate and the defect detection rate (step S404) in which the other value increases in proportion to the increase in the other value. Therefore, in the regression analysis in step S404, first, a plurality of teacher projects are selected. This teacher project selection is performed by the teacher project selection unit 11 using the development rank and the data / logic ratio.

  Next, the defect detection rate of the teacher project is calculated with reference to the defect data (FIG. 3) of the selected teacher project. Further, the coverage rate of the teacher project is calculated with reference to the review / test data (FIG. 24). In this way, a combination of (coverage rate and detection rate) is obtained for each project, and plotted in two-dimensional coordinates.

FIG. 26 is a diagram for explaining regression analysis of the coverage rate and the detection rate.
The left figure of FIG. 26 is the figure which plotted the coverage rate and detection rate calculated | required for every project on the two-dimensional coordinate. Here, a regression line shown in the right diagram of FIG. 26 is obtained by linear regression analysis. The least square method is used as a method of linear regression analysis. The least square method is a method of calculating a straight line that minimizes the sum of the squares of the distances to the respective points. Using this calculated regression line, the detection rate can be estimated from the coverage rate. For example, as shown in the right diagram of FIG. 26, the detection rate FR when an 80% coverage rate is set as an improvement measure can be obtained.

  On the other hand, the review time, defect detection rate (step S406), coverage rate, test design man-hours (step S402), and coverage are related to the increase in one value but the other value does not decrease but the increase rate decreases. Since the relationship between the rate and the number of test execution steps (step S402) is not a linear relationship, the above-described linear regression cannot be directly applied. Therefore, linear regression is performed after taking the logarithm of a numerical value with a large increase.

  First, select multiple teacher projects. This teacher project selection is performed by the teacher project selection unit 11 using the development rank and the data / logic ratio. Next, the combination of each numerical value is calculated | required using the data of each teacher project. For the review time and the defect detection rate, the review time can refer to the review / test data (FIG. 24), and the defect detection rate can refer to the defect data (FIG. 3). For the coverage rate and the test design man-hour, both values are obtained by referring to the review / test data (FIG. 24). For the coverage rate and the number of test execution steps, both values are obtained with reference to the review / test data (FIG. 24). Next, the logarithm of a numerical value with a large increase is taken and plotted in two-dimensional coordinates.

  If it is the review time and the defect detection rate, it will be (log (review time), defect detection rate), if it is the coverage rate and test design man-hours, it will be (log (test design man-hour), coverage rate). For man-hours, (log (test execution man-hour), coverage rate).

  Thereafter, the linear regression analysis by the least square method is performed as in the method shown in FIG. Thereby, if it is review time and defect detection rate, defect detection rate can be estimated from review time. If the coverage rate and the test design man-hour are sufficient, the test design man-hour can be estimated from the coverage rate. If the coverage rate and the test execution man-hour are sufficient, the test execution man-hour can be estimated from the coverage rate.

  In step S509 of FIG. 27, the value calculated in this way is reflected in the process flow model. As a result, the total man-hours after the implementation of the measure can be evaluated. The reflection method is described below.

If it is an improvement measure for the review, the increase in the review time and the increase in the defect detection rate can be known. Consider the time when detailed design review measures were taken. First, an increase in the review time is added to T3. Next, the effect of the defect is reflected. The defect detection rate is set to an estimated value (referred to as FR) so that the defect total ratio and the defect discovery ratio between the processes after the review process do not change. That is, assuming that the number of defects bi with “′” is the number of defects after the improvement measure is reflected, the values of b33 ′, b43 ′, b53 ′, b63 ′, b73 ′, and b83 ′ can be calculated using the following equations. . The following formula is an example of a detailed design review measure.
b33 '= FR × (b33 + b43 + b53 + b63 + b73 + b83)
b43 '= {(b43) / (b43 + b53 + b63 + b73 + b83)} × {b33 + b43 + b53 + b63 + b73 + b83-b33'}
b53 '= {(b53) / (b43 + b53 + b63 + b73 + b83)} × {b33 + b43 + b53 + b63 + b73 + b83-b33'}
b63 '= {(b63) / (b43 + b53 + b63 + b73 + b83)} × {b33 + b43 + b53 + b63 + b73 + b83-b33'}
b73 '= {(b63) / (b43 + b53 + b63 + b73 + b83)} × {b33 + b43 + b53 + b63 + b73 + b83-b33'}
b83 '= {(b83) / (b43 + b53 + b63 + b73 + b83)}
According to this equation, b33 ′ + b43 ′ + b53 ′ + b63 ′ + b73 ′ + b83 ′ = b33 + b43 + b53 + b63 + b73 + b83 is established. That is, the total number of defects is the same.

  If it is an improvement measure for the test, the increase in test execution man-hours, the increase in test design man-hours, and the increase in defect detection rate can be found. Assume that you have taken test measures for integration testing. First, the increase in test execution time and the increase in test design time are reflected. The test man-hour is added to T6. Assuming that the test design of the coupling test is performed by the basic design, the increase in the test design man-hour is added to T2.

  Next, the defect matrix (FIG. 8) is used to reflect the influence on the number of defects. At that time, the defect detection rate is set to an estimated value (referred to as FR) so that the total defect number and the defect detection ratio between the processes after the integration test process do not change. That is, assuming that the number of defects bi with “′” is the number of defects after reflection of the improvement measure, the values of b62 ′, b72 ′, and b82 ′ can be calculated using the following equations. The following formula is an example of a detailed design review measure.

b62 '= FR × (b62 + b72 + b82)
b72 '= {(b72) / (b72 + b82)} x {b62 + b72 + b82-b62'}
b82 '= {(b82) / (b72 + b82)} x {b62 + b72 + b82-b62'}
According to this equation, b62 ′ + b72 ′ + b82 ′ = b62 + b72 + b82 is established. That is, the total number of defects is the same.

  Then, using the number of defects bi 'after reflecting the improvement measure, the process flow model Ti, ti, and total man-hour are updated to the values after the process improvement measure according to the above-described procedure.

As described above, by setting the review time and coverage rate, it is possible to refer to past project data and evaluate process improvement measures.
That is, in step S506, the result output unit 16 displays the evaluation result on the monitor of the input / output device 2. The contents to be displayed here are the total man-hours, Ti and ti values of the process flow model, defect matrix, number of defects remaining in the subsequent processes, review time for process improvement measures, coverage rate, test design man-hours, test execution man-hours, It is the total value of the total man-hours for each process.

  From this evaluation result, the user confirms, for example, whether or not the total value of the total man-hours of each process has decreased compared to before the measure and the degree of the decrease. If the total number of man-hours for each process is reduced, the test content has been enhanced and the number of defects has been reduced, so development with higher quality and accuracy can be achieved with less man-hours. It becomes possible.

  In the case of Yes in step S507, that is, when the evaluation is finished, the operation of the process improvement measure evaluation device 1 is finished. The user adopts the man-hours and the review time allocated to each process obtained here as the plan values for promoting the project. In the case of No in step S507, that is, when the evaluation is continued, after executing steps S508 and S509, the process returns to step S506 to perform the evaluation again.

[Effect of the embodiment]
In the embodiment described above, the initial process and the return process are introduced into the project development process as separate processes, and a model that grasps the man-hour required for each process is created. As a result, it is possible to evaluate in advance and quantitatively the impact of implementing process improvement measures. In addition, the accuracy of evaluating process improvement measures can be increased.

  Note that the functions described in the above-described embodiments are not limited to being configured using hardware, but can be realized by causing a computer to read a program describing each function using software. Each function may be configured by appropriately selecting either software or hardware.

Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage.
Various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  DESCRIPTION OF SYMBOLS 1 ... Process improvement measure evaluation apparatus, 2 ... Input / output device, 10 ... Measure information input part, 11 ... Teacher project selection part, 12 ... Process flow model part, 13 ... Defect information extraction part, 14 ... Man-hour information extraction part, 15 ... man-hour calculation part, 16 ... result output part 17 ... measure influence estimation part, 20 ... control part, 21 ... man-hour DB, 22 ... defect DB, 23 ... project DB, 24 ... review test DB.

Claims (5)

For each process of the project, a process flow model section that creates a model that defines the number of man-hours when there is no defect, the number of defects to be corrected in this process, and the man-hour required per defect in this process,
A setting unit for setting the number of defects and man-hours for each of the processes from the teacher project in which the actual results have been recorded in the past,
Change the number of defects in each process set in the model based on the process improvement measure information that changed the planned value of the project, and update the man-hours of each process after the process improvement measure based on the changed number of defects The measure impact estimation unit,
A process improvement measure evaluation apparatus comprising: an output unit that outputs a process improvement measure evaluation value including the updated information.   The process improvement measure evaluation apparatus according to claim 1, wherein the process improvement measure evaluation value output by the output unit includes a total value of the total man-hours of the respective steps.   The process improvement measure according to claim 1, further comprising a selection unit that selects the teacher project based on a data / logic ratio that is a ratio of a data development amount and a logic development amount in a software development project. Evaluation device. The setting unit
If the process of the teacher project does not record the number of man-hours when there is no defect and the number of man-hours required for each defect, from the discovery of each defect to the confirmation of correction for each process The process improvement measure evaluation apparatus according to claim 2 or 3, wherein the man-hour when the defect is not present and the man-hour required for each defect are calculated based on the period information and the number of defects generated. . For each process in the project, create a model that specifies the man-hours when there are no defects, the number of defects to be corrected in this process, and the man-hours required for each defect in this process,
Set the number of defects and man-hours for each of the above steps from the teacher project whose results have been recorded in the past in the model,
Change the number of defects in each process set in the model based on the process improvement measure information that changed the planned value of the project,
Update the man-hours for each process after the process improvement measures based on the changed number of defects,
A process improvement measure evaluation method characterized by outputting a process improvement measure evaluation value including the updated information.

Images