JP5749214B2 - Bug density index value calculation device and bug density index value calculation method - Google Patents

Description

  The present invention relates to a technique for calculating a bug density index of software whose functions have been expanded.

  In order to quantitatively perform quality control during software development, the development scale, number of test items, number of bugs detected, time from test start to bug detection (number of test items used), etc. are measured, and the actual values are constant. It has been widely practiced to determine the security of software quality by reaching the target value (index value).

  Terms such as estimated residual defect density, defect detection density, test item density, defect convergence rate, and test coverage rate have been known as index values, but the calculation method for the values is not clear, and each software development It is thought that it depends on the know-how of individuals and organizations.

  One of index values that are often used is bug density (number of bug records / software development scale). This is based on the premise that the density of bugs created during the development of software with similar language and processing content is roughly constant, and based on the past bug density actual values during the development of similar software, It defines the bug density index value. For example, in the case of software with expanded functions, the bug density index value at the time of development of the next version is centered on the average value of the past bug density results of several previous versions, and the tolerance of variation is fixed at ± several percent Use the value plus the value.

JP-A-4-100161 JP 7-146806 A JP 2000-56961 A

  However, in the above method, it is difficult to clarify the basis such as which range of the past average value is used, and what percentage of the allowable range of variation is to be used. Will require know-how based on

  The present invention has been made in view of the above, and an object of the present invention is to calculate a quantitative index value for judging quality assurance of software whose functions have been expanded.

  The bug density index value calculation apparatus according to the first aspect of the present invention comprises a storage means for storing a bug density actual value obtained by dividing a bug actual number by a development scale for each version of software, and the bug density actual value from the storage means. An approximate value calculation means for reading and exponentially approximating and calculating a bug density approximate value for each version of the software, and a logarithmic value obtained by taking a logarithm of variation obtained by dividing the bug density actual value by the bug density approximate value. Variation tolerance value calculating means for calculating the variation allowable minimum value and the variation allowable maximum value from the average and standard deviation of the logarithmic value, and the bug density approximation value of the latest version of the software to the next version Bug density index value calculating means for calculating a bug density index value by multiplying the difficulty level evaluation value by the variation allowable minimum value or the variation allowable maximum value.

  In the mounting of the bug density index value calculation, the bug density actual value is stored for each function part of the software and for each test process, and the bug density index value is calculated for each function part of the software and for each test process. Features.

  The bug density index value calculation method according to the second aspect of the present invention reads the bug density actual value from the storage means for storing the bug density actual value obtained by dividing the number of bug actuals by the development scale for each version of the software, and approximates the index. Calculating a bug density approximate value for each version of the software, and calculating a logarithmic value obtained by logarithm of variation obtained by dividing the bug density actual value by the bug density approximate value for each version of the software, A step of calculating a variation allowable minimum value and a variation allowable maximum value from an average and standard deviation of logarithmic values; and a bug density approximation value of the latest version of the software, a difficulty evaluation value of the next version and a variation allowable minimum value or And calculating a bug density index value by multiplying the maximum allowable variation.

  According to the present invention, it is possible to calculate a quantitative index value for determining quality assurance of software whose functions have been expanded.

It is a functional block diagram which shows the structure of the bug density index value calculation apparatus in this Embodiment. It is a flowchart which shows the flow of a process of the bug density index value calculation apparatus in this Embodiment.

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

  FIG. 1 is a functional block diagram showing a configuration of a bug density index value calculation apparatus according to the present embodiment.

  The bug density index value calculation device 1 shown in FIG. 1 includes a bug density actual value calculation unit 11, a bug density approximate value calculation unit 12, a variation allowable value calculation unit 13, a bug density index value calculation unit 14, and a storage unit 15. . Each unit included in the bug density index value calculation device 1 may be configured by a computer including an arithmetic processing device, a storage device, and the like, and the processing of each unit may be executed by a program. This program is stored in a storage device included in the bug density index value calculation apparatus 1, and can be recorded on a recording medium such as a magnetic disk, an optical disk, or a semiconductor memory, or provided through a network. Hereinafter, each part will be described.

  The storage unit 15 stores the development scale, the number of bug achievements, the bug density achievement value, and the difficulty level evaluation value in each version of the software whose functions have been expanded.

  Development scale indicates the scale of software. In the present embodiment, the calculation is performed for each software function unit (for example, a call processing function unit, a maintenance function unit, etc.).

  The actual number of bugs indicates the number of detected software defects. In the present embodiment, the measurement is performed for each test process (for example, a binding test, a comprehensive test, etc.) and for each functional unit.

  The bug density result value is a value obtained by dividing the number of bug results by the development scale. Correspond to the functional part / test process divided by the number of bugs.

  The difficulty level evaluation value is a numerical value reflecting the software development difficulty level. Expect from the results of interviews with software developers. Set according to the functional unit / test process divided by the number of bugs. In the present embodiment, the difficulty evaluation value of the first version is set to 1, and is corrected to + when development is difficult, and to-when easy. A specific correction value can be obtained from experience.

  The bug density actual value calculation unit 11 reads the bug actual number and the development scale from the storage unit 15 for each software version, calculates the bug density actual value by dividing the bug actual number by the development scale, and calculates the bug density actual value. The data is stored in the storage unit 15.

  The bug density approximate value calculation unit 12 exponentially approximates the bug density actual value stored in the storage unit 15 to calculate the bug density approximate value for each version of the software. Versions with a bug density actual value of 0 are excluded from the values that are the basis of approximate calculation.

  The variation allowable value calculation unit 13 calculates a variation allowable minimum value and a variation allowable maximum value from the average and standard deviation of logarithmic values obtained by taking the logarithm of variation obtained by dividing the bug density actual value by the bug density approximate value. Specifically, for each software version, the actual bug density value is read from the storage unit 15 and divided by the bug density approximate value calculated by the bug density approximate value calculation unit 12 to calculate the variation, and the logarithm of the variation is calculated. Above, the mean μ and standard deviation σ are calculated based on the logarithmic values of all the versions (except for the version with 0 bug results). A value obtained by calculating the index of (μ−σ) is set as a minimum allowable value for variation, and a value calculated by the index of (μ + σ) is set as a maximum allowable value for variation. Instead of ± σ, the variation width may be ± 2σ or ± 0.5σ, for example. Whether to widen or narrow the width of variation can be obtained from experience. In this embodiment, ± σ is used as a standard. Note that the base for calculating the logarithm and the exponent is the same, for example, the base of the natural logarithm.

  The bug density index value calculation unit 14 multiplies the value obtained by multiplying the bug density approximate value of the latest developed version calculated by the bug density approximate value calculation unit 12 by the difficulty evaluation value of the next version of the latest version. Multiply the minimum value by calculating the minimum bug density index allowable value in the next version, and multiply by the maximum variation allowable value to calculate the maximum bug density index allowable value in the next version.

  Further, the input unit 2 inputs the development scale, the number of bug results, and the difficulty level evaluation value for each version and for each functional unit / test process, and stores them in the storage unit 15. The bug density actual value is calculated by the bug density actual value calculation unit 11 and stored in the storage unit 15.

  The output unit 3 outputs a bug density index value (bug density index allowable minimum value, bug density index allowable maximum value) for each functional unit / test process in the next version calculated by the bug density index value calculating unit 14. The quality assurance of software is judged by the fact that the bug density actual value of the next version falls within the range from the bug density index allowable minimum value to the bag density index allowable maximum value.

  Next, the process flow of the bug density index value calculation apparatus 1 will be described.

  FIG. 2 is a flowchart showing a processing flow of the bug density index value calculation apparatus 1.

  First, the bug density record value calculation unit 11 reads the number of bug records and the development scale of each version from the storage unit 15, and calculates the bug density record value X (i) by dividing the bug record number by the development scale (step S11). . Here, i indicates the version number and satisfies m (the oldest version) ≦ i ≦ n (the latest version that has been developed). Note that m = 1 is not necessarily required.

  Subsequently, the bug density approximate value calculation unit 12 calculates the bug density approximate value X ′ (i) of each version by exponential approximation based on the bug density actual value X (i) in the range of m ≦ i ≦ n (step) S12). When calculating the bug density approximate value, the version j in which the number of bug results is 0, that is, the version j in which X (j) = 0 is excluded. It is also possible to know the bug density approximate value X ′ (j) of the version j that has been excluded.

  Further, the variation allowable value calculation unit 13 calculates the variation C (i) = X (i) / X ′ (i) in the range of m ≦ i ≦ n (step S13), and the logarithmic value ln of the variation C (i). The average μ and standard deviation σ of [C (i)] are calculated (step S14). At this time, the version with the bug record number 0 is excluded.

  Then, a variation allowable minimum value Cmin (n + 1) and a variation allowable maximum value Cmax (n + 1) in the next version n + 1 are calculated (step S15). The variation allowable minimum value Cmin (n + 1) and the variation allowable maximum value Cmax (n + 1) are calculated by Cmin (n + 1) = exp (μ−σ) and Cmax (n + 1) = exp (μ + σ), respectively.

  Finally, the bug density index value calculation unit 14 calculates the bug density approximate value X ′ (n) of the latest version n, the difficulty evaluation value D (n + 1) of the next version n + 1, the variation allowable minimum value Cmin (n + 1), and the variation allowable maximum The bug density index allowable minimum value Zmin (n + 1) and the bug density index allowable maximum value Zmax (n + 1) of the next version n + 1 are calculated from the value Cmax (n + 1) (step S16). The bug density index allowable minimum value Zmin (n + 1) and the bug density index allowable maximum value Zmax (n + 1) are respectively Zmin (n + 1) = X ′ (n) × D (n + 1) × Cmin (n + 1), Zmax (n + 1) = Calculation is performed by X ′ (n) × D (n + 1) × Cmax (n + 1).

  Next, the reason why exponent approximation can be used for the actual bug density value will be described.

  In the case of software that has been expanded in function, even if the software developers are different, the organization that took over the know-how generally repeats horizontal deployment and pausing for each detected bug, so know-how It is considered that the detected bug density in the test process shows a decreasing trend as the version number progresses on average. However, there is a difference in the difficulty of development contents depending on the version, and it is considered that the bug density increases in difficult development and the bug density decreases in easy development.

  Here, the ratio of the bug density actual value X (i + 1) in the next version i + 1 to the bug density actual value X (i) in a certain version i is defined as D (i + 1) = X (i + 1) / X (i). , X (i + 1), X (i + 1) = X (i) × D (i + 1). When X (0) when i = 0 is set as a constant A, X (1), X (2),..., X (i) are as follows.

X (1) = X (0) × D (1) = A × D (1)
X (2) = X (1) × D (2) = A × D (1) × D (2)
・
・
X (i) = X (i-1) * D (i) = A * D (1) * D (2) ... * D (i)

When D (i) is always a constant value D, X (i) = A × D ⁱ , that is, a bug density actual value when D (i) different for each plate is approximated by the geometric mean value D X (i) can be approximated by an exponential function. When 0 <D <1, X (i) approaches 0 as the plate advances.

  The meaning of D (i) can be interpreted as corresponding to the difficulty level of development contents for each version. When D (i + 1) = 1, X (i + 1) = X (i), that is, version i + 1 is version i. It is not difficult or easy compared to, and it can be understood that the bug density is the same. In the case of software whose functions are repeatedly expanded, it can be observed that D is a value close to 1 and smaller than 1, for example, about 0.9.

  Next, calculation of the tolerance for variation will be described.

  Since the bug density actual value can be approximated by an exponential function, it can be considered that the variation of the bug density actual value for each version follows a log normal distribution instead of a normal distribution. The variation C (i) of the bug density actual value X (i) with respect to the bug density approximate value X ′ (i) in the version number i is C (i) = X ′ (i) / X (i). Since the variation C (i) is considered to follow a lognormal distribution, the average μ and standard deviation σ of the logarithmic value ln [C (i)] of C (i) is calculated, and exp (μ−σ) varies. The minimum allowable value, exp (μ + σ), can be considered as the maximum allowable variation.

  In statistics, in normal distribution, the frequency of the sample that falls within the range indicated by mean μ and standard deviation σ is about 2/3 of the whole in μ ± σ and about 19/20 of the whole in μ ± 2σ. It has been known. When the allowable width is made wider or narrower, the value on which the allowable width is based is not ± σ but, for example, ± 2σ or ± 0.5σ.

  Next, the method according to the present invention and the conventional method that has been generally used will be compared with respect to the method for calculating the bug density index value of software that is expanding functions.

  In the conventional method, a value obtained by adding a fixed value of ± several percent as an allowable variation width is used as the bug density index value centering on the average value of the actual bug density values of several past versions. The conventional method will be specifically described below.

  The average value X ″ (n) of the past k version bug density results is expressed by the following equation.

  X ″ (n) = (X (n−k + 1) +... + X (n))) / k

  When the tolerance of variation is ± ε, the bug density index allowable minimum value Ymin (n + 1) and the bug density index allowable minimum value Ymax (n + 1) in the next version n + 1 are expressed by the following equations.

Ymin (n + 1) = X ″ (n) × D (n + 1) × (1-ε)
Ymax (n + 1) = X ″ (n) × D (n + 1) × (1 + ε)

  In the conventional method, there is no theoretical basis for how k is taken, and in many cases, for example, “5” is determined in consideration of simplicity. Generally, since the bug density actual value X (i) tends to decrease as the version number i progresses, the larger the k is, the higher the bug density index value is calculated. Conversely, the smaller k is, the closer to the latest bug density actual value is calculated. In that case, if the deviation from the average of the latest bug density actual values of the latest version is large, for example, the latest version When the difficulty level is extremely difficult or easy, it is greatly affected by the deviation.

  In the method according to the present invention, since the bug density is calculated by exponential approximation based on the version whose bug density actual value is known, the above problem is solved.

  Further, in the conventional method, there is no theoretical basis for how to set ε, and in many cases, for example, “30%” is determined in consideration of past know-how and simplicity. Whether the set ε is valid or not needs to be reviewed after the next version results. In addition, since the variation of the bug density actual value follows a log normal distribution rather than the normal distribution, setting a variation allowable width equally large and small around the arithmetic mean results in a strict allowable value on the maximum value side. On the contrary, if the allowable range is set in accordance with the maximum value side, the allowable minimum value may be a negative value (a meaningless value).

  The method according to the present invention uses a statistical method. For example, when Cmin (n + 1) = exp (μ−σ) and Cmax (n + 1) = exp (μ + σ), the actual bug density value of the next version is In the state where the difficulty level is not evaluated, it can be expected that the probability is within the allowable range of the bug density index with a probability of about 2/3. Further, the allowable minimum value does not become a negative value.

  As described above, according to the present embodiment, the bug density approximate value calculation unit 12 exponentially approximates the bug density actual value to calculate the bug density approximate value, and the variation allowable value calculation unit 13 determines the bug density actual value. Is divided by the approximate value of the bug density to determine the variation of each version, the minimum allowable variation and the maximum allowable variation are calculated from the average and standard deviation of the logarithm of the variation. When developing the next version of software that has expanded its functions by calculating the bug density index value of the next version by multiplying the density approximation value by the difficulty evaluation value of the next version, the minimum allowable variation, and the maximum allowable variation, respectively. It is possible to calculate a quantitative index value for judging the quality assurance of software in the test process. The calculation of the index value according to the present invention does not require know-how for calculation and can be expected to be more accurate than the method generally used conventionally.

DESCRIPTION OF SYMBOLS 1 ... Bug density index value calculation apparatus 11 ... Bug density actual value calculation part 12 ... Bug density approximate value calculation part 13 ... Variation tolerance value calculation part 14 ... Bug density index value calculation part 15 ... Memory | storage part 2 ... Input part 3 ... Output Part

Claims (4)

A storage means for storing the bug density actual value obtained by dividing the number of bug results by the development scale for each version of the software,
An approximate value calculation means for reading the bug density actual value from the storage means and performing exponential approximation, and calculating a bug density approximate value for each version of the software;
A logarithmic value obtained by taking the logarithm of variation obtained by dividing the bug density actual value by the bug density approximate value is calculated for each version of the software, and the variation allowable minimum value and variation allowable maximum value are calculated from the average and standard deviation of the logarithmic value. Variation tolerance calculation means for calculating
A bug density index value calculating means for calculating a bug density index value by multiplying the bug density approximate value of the latest version of the software by the difficulty evaluation value of the next version and the variation allowable minimum value or the variation allowable maximum value;
A bug density index value calculation apparatus comprising:   2. The bug according to claim 1, wherein the bug density actual value is stored for each function part and test process of the software, and the bug density index value is calculated for each function part and test process of the software. Density index value calculation device. The step of reading the bug density actual value from the storage means for storing the bug density actual value obtained by dividing the number of bug actuals by the development scale for each version of the software, exponentially approximating, and calculating the bug density approximate value for each version of the software When,
A logarithmic value obtained by taking the logarithm of variation obtained by dividing the bug density actual value by the bug density approximate value is calculated for each version of the software, and the variation allowable minimum value and variation allowable maximum value are calculated from the average and standard deviation of the logarithmic value. A step of calculating
Multiplying the bug density approximate value of the latest version of the software by the difficulty evaluation value of the next version and the variation allowable minimum value or the variation allowable maximum value to calculate a bug density index value;
A bug density index value calculation method characterized by comprising:   4. The bug according to claim 3, wherein the bug density actual value is stored for each function part and test process of the software, and the bug density index value is calculated for each function part and test process of the software. Density index value calculation method.

Images