http://www.csai.cn/tszl/xmgj.htm
The scale estimate of the software project has always been more complicated, because the complexity of the software itself, the lack of historical experience, the lack of estimation tools, and some people are wrong, resulting in the size estimation of software projects and the actual situation. Therefore, it is estimated that the error has been included in the four causes of the failed software project failure.
Software engineers often ask, what kind of software is needed, how much is it takes? In the face of this problem, there are many people very difficult, because the needs of the first user is too uncharacter, second, lack a scientific estimation method. Here you introduce the estimation methods of several software project scale.
Concept introduction
First introduce a concept -Loc (line of code) that measures the size of the software project --Loc (line of code), LOC refers to all executable source line numbers, including the deliverable work control language (JCL: Job Control Language) statement, data definition , Data type declaration, equivalent statement, input / output format declaration, etc. The value of one-generation code line (1LOC) and the number of monthly months can reflect the production capacity of a software production organization. Organization can account for a single line code value of the organization based on the audit of historical projects.
For example, a software company statistics found that the source file (.c and .h file) formed by the company every 10,000 line C language source code is about 250K. The source file size of a project is 3.75m, and it is estimated that the project source code is approximately 150,000, and the project has a total investment workload of 240 months. The monthly fee is 10,000 yuan (including per capita salary, welfare, office fees) Belt, etc.), the value of 1LOC in the project is:
(240 × 10000) / 150000 = 16 yuan / LOC
The number of people who changed the project is:
150000/240 = 625LOC / Month
Method 1, Delphi method
The Delphi method is the most popular expert assessment technology. In this way, this approach is suitable for assessment between the past and future, new technologies and specific procedures, but the expert "special" and understanding of the project The degree is the difficult point in work, although Delphi technology can alleviate this deviation, expert assessment technology is usually used in the actual cost of a new software, but this method is especially useful to determine the input of other models. Delphi method encourages participants to discuss each other. This technology requires participation in various software-related experience, and persuades each other.
The step of the Delphi method is:
1. The coordinator provides project specifications and estimates to each expert; 2. Coordinator converges the team members to discuss and scale related factors; 3. Experts have anonymously filled out the representatives; 4, the coordinator sorted out an estimate summary, The representative of the representative returned to the expert; 5. The coordinator convene group will discuss a large estimate difference; 6. Expert review estimates summarize and submit another anonymous estimate on the representative; 7, repeat 4-6 until the lowest And the highest estimate.
Method 2
The sub-method is suitable for assessing some similar projects with historical projects in the application, environmental and complexity, and is estimated by the comparison of new projects and historical projects. The accuracy of the estimated results of the spectrum depends on the integrity and accuracy of the historical project data. Therefore, one of the prerequisites for use in goodbyctions is to organize the establishment of a better project evaluation and analysis mechanism, and the data analysis of historical projects. It is trustworthy.
The basic steps are: 1, organize the list of project functions and implement each function, and identify the same points and different points of each function list and historical project, especially pay attention to the historical project; 3. The estimation value of each function is obtained by steps 1 and 2; 4, a scale estimate is generated.
The use of categories in software projects often resolve the estimation of reusable code. The best way to estimate the amount of reusable code is to examine the existing code in detail by programmer or system analyst. Estimated the code percentage of code to be redesigned in new projects to reuse, and re-coding or modifying code percentages and The percentage of code required to retest. According to these three percentages, the following calculation formula is calculated: equivalent code rows = [(redesign% re-encoding% re-test%) / 3] × Existing code row
For example: there are 10,000 lines of code, assuming 30% needs to be redesigned, 50% need to be re-coded, 70% requires retest, then its equivalent code line can be calculated:
[(30% 50% 70%) / 3] × 10,000 = 5,000 equivalent code line.
即 即: Reuse this 10,000 code is equivalent to writing a workload of 5000 code lines.
Method 3, function point estimation method
Function point measurement is a size estimation method based on system function during the demand analysis phase. By studying initial application requirements to determine the number and characteristics of various input, output, calculation, and database requirements. The usual step is:
1. Calculate input, output, query, master file, and number of interface requirements. 2. Take these data. The table below is a typical weight list.
Function Type Weight Input 4 Output 5 Query 4 Main Control File 10 Interface 10
3. The estimator can be adjusted with 25%, 0, or -25% according to the judgment of complexity.
It is found that the development and function point of a software product is very helpful to the scale of the project. However, after understanding the product, the function point can be converted to the software scale measurement more commonly used LOC.
Method 4 PERT estimation method
Pert is estimated to estimate the completion time of each project activity: a product's expectation, a minimum possible estimate, a maximum possible estimate. Use these three estimates to get Pert statistical estimates for a product expected size and standard deviation. Pert estimation can obtain the expected value E of the code line E, and the standard deviation SD.
Detailed estimation method, the reader can refer to the "Application PERT Estimation Estimation" written by the author, and details will not be described here.
