JPH05173852A - Method and device for estimating reliability of software - Google Patents

Abstract

PURPOSE:To eliminate such an evil that an unreasonable analyzed result in the viewpoint of the evaluation of the reliability of software is presented, and to select the regression model of better adaptation, and to display the result quantitatively. CONSTITUTION:Plural techniques to calculate the best fit coefficients of the regression model are provided, and program routine to select the regression model of the better adaptation by excluding the unreasonable result of regression analysis by a first deciding stage 53, and simultaneously, selecting the regression model reasonable for the accumulation curve of a bug obtained from an acquired bug data group, and besides, comparing error functions quantitatively by a second deciding means 55 is incorporated. The reasonable analyzed result can be presented for various kinds of the software, and the reliability is improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a reliability estimation method and apparatus in software quality control.

[0002]

2. Description of the Related Art In recent years, a reliability estimation method using a bug growth curve has been widely used in quality control of software represented by a computer program. This technique is convenient because it is possible to scientifically grasp the quality situation by regression analysis of the time series of bugs obtained in the software development and testing process.

First, a conventional example will be described with reference to FIGS. 1 and 4. In FIG. 1, 1 is an information processing device, 2 is a computer (with FDD), 3 is a display unit,
Reference numeral 4 is a printer, 5 is a keyboard, and 6 is a floppy disk. The floppy disk 6 in which the program routine is stored is inserted into the computer (with FDD) 2 and the operation software is started to start the information processing apparatus 1. The program routine is loaded from the floppy disk 6 and is in the input waiting state.
FIG. 4 is a flow chart for explaining the conventional example, and its operation is shown below. Step 50 is a key input of bug acquisition data from the keyboard 5, for example, a step of continuously inputting the number of bugs obtained every day. Step 51 is a processing step of performing regression analysis on these data groups and calculating a best fit coefficient of a regression model (Gompertz curve, logistic curve, etc.). Step 57 is the step of displaying the cumulative curve of the data group and the regression analysis result on the same screen, and step 58 is the step of printing out the graphical display contents and the required items obtained by the data processing. The display unit 3 corresponds to the hardware in step 57, and the printer 4 corresponds to the hardware in step 58.

[0004]

As described above, in the conventional software reliability estimation method and apparatus,
Although a regression model called the best fit was always sought, it did not always show a good fit to the cumulative curve of the acquired data set. Moreover, even if an analysis result that rarely meets expectations is output, it is also a problem that no countermeasure can be taken.

The present invention has been made in order to solve the above problems, and takes measures to prevent the output of analysis results against expectations, and to quantify the deviation between the cumulative curve and the regression model of the acquired data group. Intended for display.

[0006]

A software reliability estimation method and apparatus according to the present invention prepares a plurality of calculation methods of best fit coefficients in the same regression model,
The index is compared and selected to select an appropriate regression model, and the deviation between the curves is compared and judged by an error function, and the one with better suitability is selected.

[0007]

[Function] Selecting regression models having different calculation methods and having excellent suitability makes it possible to output a wide range of appropriate analysis results for a group of field data showing various shapes.

[0008]

【Example】

Example 1. 1, 2 and 3 are diagrams for explaining an embodiment of a software reliability estimation method and apparatus according to the present invention. In FIG. 1, 1 is an information processing apparatus and 2 is a computer (with FDD). 3 is a display unit, 4 is a printer, 5 is a keyboard, 6 is a floppy
It is a disc. The hardware configuration of the present invention is the same as that of the conventional example, the floppy disk 6 storing the program routine is inserted into the computer (with FDD) 2, the operation software is activated, and the information processing apparatus 1 is started. .. The program routine is loaded and waits for input. When a key input is made from the keyboard 5, the program routine is operated to display the processing result chart on the display unit 3 and to print out the processing result and required items on the printer 4.

Here, a so-called regression analysis will be briefly described. FIG. 3 shows an example of the analysis result.
As is well known, debugging work is performed in the software development / testing process. This is a bug removal work to improve the quality of software. It takes time or period on the horizontal axis x and acquires on the vertical axis y. The figure that takes the cumulative total number of bugs that are called is called the bug growth curve. Suppose that the cumulative curve of the acquired bug data is G (x), the process proceeds from x ₀ to x _{1 for the} first time, and the bug of y ₁ is obtained at this time. Curve G (x)
It can be seen that a large number of is similar to the unique mathematical expression, but this unique mathematical expression is called a regression model. In the equation F (x), which is unique for any coefficient, the curve G
Regression analysis determines the coefficient of F (x) that best fits (x). Saturation curves are usually used for regression models because bugs will disappear someday. F (x)
If is determined converged by x _m, a F (x _m) K
Is the total number of predicted bugs, and L, which is the difference between K and y ₁ , is the number of remaining bugs. Further, it can be seen that the date of the difference between x _m and x ₁ is further required to detect all the bugs. Now, in rare cases, the cumulative total number of bugs y ₁ may exceed the predicted total number of bugs K with respect to the above-described valid analysis result. Especially when the cumulative total number of bugs y ₁ is relatively small,
It is seen in the case of a plurality of bump-shaped cumulative curves and the like, which is not preferable as a predicted value.

The reliability of the regression analysis is premised on the selection of a regression model having a higher degree of conformity. The selection method of the present invention will be described below. The Gompertz function, which is a typical regression model, is given by the following equation.

[0011]

[Equation 1]

In Expression 1, a and b are coefficients calculated from sample data, and K is the total number of prediction bugs calculated using all data. Here, paying attention to the fact that the method of calculating the differential coefficient in the regression analysis is an approximate expression, the following two expressions are defined.

[0013]

[Equation 2]

[0014]

[Equation 3]

Equation 2 corresponds to the method using two sample points for the above-described calculation of the differential coefficient, and Equation 3 corresponds to the method using three sample points. The regression model obtained using equation 2 F _A (x), the predicted bug total number and K _A, the regression model was determined using the number 3 F _B (x), the predicted bug total number K _B
And Also, the cumulative curve G (x) of the bug and the regression model F
The following function is defined using (x).

[0016]

[Equation 4]

Equation 4 will be referred to as an error rate. Here, x ₀ and x ₁ are integration intervals. The error rate ε is, so to speak, 2
The deviation of two functions F (x) and G (x) is quantitatively displayed, and becomes zero when the two functions overlap. That is, it is considered that the smaller the error rate is, the better the fitness is. Here, the error rate when F _A (x) is used as F (x) is ε _A , and the error rate when F _B (x) is used as F (x) is ε _B.

FIG. 2 is a flow chart for explaining the embodiment of the present invention, and its operation will be described step by step. Step 50 is a key input step of the acquired bug data group, and steps 51 and 52 are the above-mentioned F _A (x), K _A , ε _A and F _B (x), K _B , ε _{B, respectively.}
Is a processing step for obtaining Step 53 is a step of comparing and judging the total number of predicted bugs K _A and K _B and the cumulative total number of bugs y ₁ , and when both K _A and K _B are larger than y ₁ , K _A or K _B.
When only _{one of them is} larger than y ₁ and when both K _A and K _B are y ₁ or less, they are classified into four types. K _A only y ₁
If it is greater than F (x) in step 54A,
_{If A} (x) is assigned and only K _B is greater than y ₁ , then F (x) is assigned F _B (x) in step 54B. If both K _A and K _B are y ₁ or less,
“Regression analysis result is not valid” in step 54C
Display comments such as. When both K _A and K _B are larger than y ₁ , the second comparison / determination step of step 55 is reached.
In this case, the magnitude relationship between ε _A and ε _B is compared, and ε _A is ε _B
If it is smaller than F (x) in step 56A,
_{If A} (x) is assigned and ε _B is less than or equal to ε _A , then F (x) is assigned to F (x) in step 56B.
_B (x) is assigned. In this way, the optimum regression model F (x) is selected, the cumulative curve of the data group and the regression analysis result are displayed on the same screen in step 57, and obtained in step 58 by graphical display contents and data processing. Print out the required items.

Example 2. In the above embodiment, the case where the Gompertz function is used as the regression model has been described, but it is also possible to use a regression model of a different type and to obtain the same effect as that of the embodiment. Further, in the above-described embodiment, the example of using two sampling points and the example of using three sampling points have been described as the calculation method of the regression model. Even when weighting is applied, the same effect as the embodiment is obtained.

[0020]

As described above, according to the method and apparatus for estimating the software reliability of the present invention, the regression results obtained by different calculation methods are used to make the analysis result valid and to be obtained from the acquired data. Since it is configured to select the one that better fits the bug accumulation curve,
A wide range of higher reliability can be guaranteed for a wide variety of software. Also, by introducing the error rate,
The degree of conformity between curves can be grasped quantitatively, and handling becomes convenient.

[Brief description of drawings]

FIG. 1 is a system configuration diagram according to an embodiment of the present invention.

FIG. 2 is a flow chart for explaining the operation of the present invention.

FIG. 3 is an explanatory diagram showing an example of an analysis result of the present invention.

FIG. 4 is a flowchart showing a conventional example.

[Explanation of symbols]

 1 Information processing device 2 Computer (with FDD) 3 Display unit 4 Printer 5 Keyboard 6 Floppy disk 50 Key input step 51 Data processing step 1 52 Data processing step 2 53 Judgment step 1 54 Selection step 1 55 Judgment step 2 56 Selection Stage 2 57 Display stage 58 Print stage

Claims (2)

[Claims] 1. In a software reliability estimation method for regression analysis of bugs generated in a software testing process and grasping the quality status using the analysis result, the method of calculating the acquired bug data group is different. Regression analysis is performed using a regression model, and the cumulative total number of acquired bugs obtained from the cumulative curve of the bug data group is compared with the predicted total number of predicted bugs obtained from the regression model with a different calculation method, and the cumulative total of the bug data group is calculated. A software reliability estimation method characterized in that a difference between a curve and a regression model having a different calculation method is compared and determined by an error function, and an optimal regression model calculation method is selected and used as a regression analysis result. 2. A display unit, a keyboard, a printer, a memory in which the following program routines are stored, and a central processing unit that executes according to the contents of this memory, and regression analysis is performed on the input data group. Then
An apparatus for estimating software reliability, characterized in that the processing result can be visually observed on the display unit and can be obtained as a print output of the printer. (A) A first program routine for performing regression analysis on the input data group using regression models of different calculation methods to obtain the best fit coefficient, the total number of predicted bugs, and the numerical deviation between the curves. (B) The cumulative total number of acquired bugs obtained from the input data group and the total number of predicted bugs obtained from each regression model are compared and determined, and a regression model is selected according to the size of the number, or A second program routine that displays a comment. (C) A third program routine for comparing and judging the numerical deviation between the curves with respect to the one satisfying the comparison condition and selecting a regression model suitable for the cumulative curve of the input data group. (D) A fourth program routine for graphically displaying the cumulative curve of the input data group and the selected regression model on the same screen. (E) The fifth program routine for printing out the contents graphically displayed on the same screen and the required items obtained by the data processing.