Software research methods

The main method of software research is measurement. To control the processes, products and resources of the software life cycle, you should use the values, characterizing their properties, called metrics. This is a certain property of the object with which you can make a value. To synthesize the value, it is necessary to determine the property (the semantics of the value), the system of values ​​(scale), and the method of comparing the values ​​with the value. In the theory of measurement, three main measurement scales are distinguished: nominal, order and quantitative. Nominal (classification) scale includes values, manifests itself only with respect to equivalence. Or it can be comparable with the property of the object (Not ordered one in relation to the other). For example, you can compare the source code of a program with the value of a language whose value can be the name of one of the languages ​​(for example, C, C + +, Pascal, Java, etc.).

The same type has a scale for assigning software modules (for example, Databases, Mathematical packages, Operating systems, etc.). To the nominal values, only the pitch operation for equivalence is applied. Ordinal scale of observation of the compilation of one value relative to another, to which the comparison operations relate.

An ordinal pick-up can be set for most expert assessments, for example, the evaluation of the readability of the program text is unsatisfactory, satisfactory, good, excellent, or for evaluating the level of encapsulation of software components, lexical, operator, procedural, class, modular. The quantitative scale includes meanings, manifested itself in relation to equivalence, order and adequacy. Such quantities allow you to forge adequate and multiplicative operations on the values ​​(subtraction, multiplication, division). These include, for example, such a number of lines of code, compiling a comment, estimating the effort to create code.


28 October 2017

You may be interested in