The paper is devoted to the developing the intelligent system for determining the sufficiency of metric information in the software requirements specifications (SRS), which provides on the basis of the natural language processing of SRS: conclusion about the sufficiency of metric information (indicators for metrics calculation) in SRS, numerical assessment of the sufficiency level of metric information in the SRS, visualization of missing indicators for metrics calculation. The developed intelligent system provides an increase in the sufficiency of information by 12.71-50.28% for the SRS of the information and analytical system for the accounting of therapeutic and diagnostic activities provided to the wounded during transportation. In general, the developed system provides an increase in the sufficiency of metric information in the SRS to 100% - if it's necessary (for critical software) or at the customer's request. The developed intelligent system for determining the sufficiency of metric information in the SRS can be used during the software development for government agencies, military formations and law enforcement agencies, commercial organizations (both for organizationsdevelopers of software and for organizations-customers of software).
Given the fact that in today's world software development has become one of the most
expensive industries, and any bottlenecks in the technological process of its creation
can lead to unwanted results, one of the main requirements of users to modern
software is its high quality and low complexity. According to ISO 25010 [
Today, there are a number of models that provide the calculation of the software
quality and complexity, but the multiplicity of interpretation of these characteristics
complicates such calculations. Most models are based on the use of different software
metrics. According to ISO 24765 [
The using of quantitative metrics had to help in solving a number of practical
tasks [
The modern software industry has accumulated a large number of metrics, that
assess the certain production and operating features of software, but modern software is
not ideal in terms of its quality [
The complexity of the justification of the choice and interpretation of metrics during making the production decisions, and ignoring the stages of the software lifecycle don't allow the full use of metrics for evaluating and predicting the software characteristics at the early stages of the software lifecycle. At present, the actual task is calculating the metrics' values for estimating and predicting the quality and complexity of software at the early stages of the lifecycle.
Let's consider the known methods and tools for evaluating the values of complexity and quality metrics.
The German National Research Center of Information Technology (GMD) has
developed the ProcePT project [
IBM corporation offers the methodology for creating complex software systems,
called Cleanroom Software Engineering [
The Logiscope package [
Pure Software company, a leading developer of automated software tools for
creating high-quality software, offers the Purify system [
The Hindsight tool of IntegriSoft [
The EzCover tool [
Metricate tool [
DMS tool [
In the paper [26] the tool is developed, named as SWMetrics, using Microsoft Visual Studio C# to compute a metrics of LOC, SLOC and complexity based the cyclomatic complexity metric for quality measurement for many format languages of the source of code.
The paper [27] discusses how the Expert System can be used to automate the selection and implementation process of software quality assurance and provides recommendations for building an expert system for software quality assurance which can be used to help software-producing organizations in selecting the most suitable models to be adopted according to their properties and needs.
Authors of the paper [28] present a software quality support tool, a Java source code evaluator and a code profiler based on computational intelligence techniques that represent a new approach to evaluate and identify inaccurate source code usage and transitively, the software product itself. The aim of this project is to provide the software development industry with a new tool to increase software quality by extending the value of source code metrics through computational intelligence.
Authors of [29] discuss software measurement and metrics and their fundamental role in the software development life cycle, with focus on software test metrics, discuss their key role in software testing process and also classify and systematically analyze the various test metrics.
The proposed in [30] BornBaby model is a whole new dimension of Artificial Intelligence for software engineering domain: attempts are being made to come up with a software synthesis program, which is able to write programs on its own, like human programmers; defines how a program must learn in order to come up with solutions; emphasizes on how the program must learn the data based on natural concepts of living beings and the implementation is generic.
A fuzzy logic reputable paradigm is proposed in [31] for predicting software defect density on individual phases of the software development lifecycle.
The thesis [32] presents methodological investigations using search-based techniques, which are relevant to the task of software quality measurement and prediction, using search-based techniques in large-scale projects during verification, validation and testing of software.
The general disadvantages of the above tools of quality assessment are: a subjective dependence of the choice of metrics that tools of automating the calculation of metric information will calculate; the subjectivity of metric interpretation, because the exact (standard) values of the metrics are absent; the focus of automated "measuring" tools is the testing and metrology of the finished source code, but not the prediction and calculation of software metrics at the design stage.
In [
Then, the task of assessing the sufficiency of metric information at the early stages of the software lifecycle, in particular, in the SRS (as the possibility of obtaining the indicators for calculation of metrics' values), is actual. Although completeness of software requirements is desirable, determination of completeness of the set of requirements is not realistic as was proved in [33]. The insufficiency of information on indicators in requirements will lead, accordingly, to the impossibility of calculating the values of certain metrics and hence to the reduction of the validity of assessments of software quality and complexity at the early stages of the lifecycle. Today the assessment of sufficiency of software safety requirements is known [33, 34].
Then, the aim of this study is the developments of the intelligent system for
determining the sufficiency of metric information in the SRS, which, based on the
natural language processing of SRS, will provide conclusions about the sufficiency of
metric information in the SRS (of indicators for calculation of the chosen in [
Intelligent system (IS) for determining the sufficiency of metric information in the SRS is developed as an agent-oriented system, which consists of two intelligent agents (Fig. 1): the intelligent agent (IA) for parsing the SRS on the search of metric information (indicators for calculation of metrics) and the IA for determining the sufficiency of metric information in the SRS.
Both intelligent agents are built on the basis of the ontological approach. IAs use the early developed base ontology of the subject domain "Software Engineering" (part "Quality and complexity of software. Metric analysis") as the known knowledge.
Fig. 1. Intelligent system for determining the sufficiency of metric information in the SRS.
The method of activity of the IA for semantic parsing the SRS on the search of metric
information (indicators for the calculation of metrics) consists of the following steps:
1. Search of each indicator of the base ontology "Software Engineering" (part
"Quality and complexity of software. Metric analysis") in the SRS for the real
software (such the base ontology is contained in the agent's knowledge base).
2. If <indicatorі> is in the SRS, then <indicatorі> belongs to the set of available
indicators, i=1..42 (since, according to [
In [35], we have developed method of activity of ontology-based intelligent agent for evaluating the initial stages of the software lifecycle, which, on the basis of comparison of the obtained real ontology with the base ontology for non-functional characteristics of the software, forms the conclusion about the sufficiency or insufficiency of the information, and numerical assessments of the sufficiency level of the SRS information for defining the non-functional characteristics. Now we use this method for development of the IA for determining the sufficiency of metric information in the SRS. But the base ontology for such IA is the ontology "Software Engineering" (part "Quality and complexity of the software. Metric analysis"). In addition, the numerical assessment of the level of sufficiency of metric information in the SRS will be calculated by the formula (1):
D
k qmi j
k
j 1 qni j
k
where – quantity of metrics ( , because in [
Thus, the developed IS for determining the sufficiency of metric information in the SRS performs the parsing (semantic analysis) of the natural language SRS with the purpose of the search of the indicators, which are necessary for calculation of the software metrics, and also forms the conclusion about the sufficiency or insufficiency of metric information in the SRS, asesses the level of sufficiency of metric information and visually shows missing indicators with the distribution by the metrics, for which they are used.
Intelligent system for determining the sufficiency of metric information in the SRS has implemented in the form of free software, which is available by the link – https://olp-project.herokuapp.com.
The user of the IS for determining the sufficiency of metric information in the SRS upload the SRS in pdf-format. IA for parsing the SRS on the search of metric information (indicators for calculation of metrics) performs parsing of the specification and generates the ontology for real software as a .owl file, which the user can download (Fig. 2).
Next, the IA for determining the sufficiency of metric information in the SRS performs a comparison of the ontology for real software with the base ontology and concludes about the sufficiency of metric information, namely: Count of missing indicators (without considering the number of uses of each indicator), Count of missing indicators (considering the number of uses of each indicator), Percent of missing indicators (without considering the number of uses of each indicator), Percent of missing indicators (considering the number of uses of each indicator), Total numerical evaluation of information sufficiency level for all metrics in SRS, and visualized list of missing indicators with distribution by the metrics.
For the experiments, two SRS of the information and analytical system for the accounting of therapeutic and diagnostic activities provided to the wounded during transportation were analyzed, which were developed by two different software companies at Khmelnitsky.
The level of sufficiency of metric information in the SRS1 was 49.72% when the 7 indicators without considering the number of uses of each indicator and 31 indicators considering the number of uses of each indicator were absent – Fig. 3.
The level of sufficiency of metric information in the SRS2 was 75.28% when the 14 indicators without considering the number of uses of each indicator and 20 indicators considering the number of uses of each indicator were absent – Fig. 4.
Obviously, the sufficiency of metric information in SRS2 is higher than the sufficiency of metric information in SRS1, and the number of missing indicators considering the number of uses of each indicator in SRS2 is lower than in SRS1 (while the number of missing indicators without considering the number of uses of each indicator in SRS2 is twice as high as in SRS1). These results are due to the fact that more important and priority are indicators that affect more than one metric.
On Fig. 5 the visualization of the missing indicators for determining the metrics of complexity with exact values at the design stage are represented with the distribution by metrics is represented. Fig. 4. Conclusion about the sufficiency of metric information in SRS2, which is provided by the intelligent system for determining the sufficiency of metric information in the SRS.
The customer of the software of the information and analytical system for the accounting of therapeutic and diagnostic activities provided to the wounded during transportation considered the level of sufficiency equal 95% is acceptable (since medical software is the critical software). So the customer demanded the re-work of both SRS by their developers. After re-work, the SRS were again analyzed by the developed IS – Fig. 6, Fig. 7.
The level of sufficiency of the metric information in SRS1 after re-work is 100% (the SRS1 has all the necessary indicators for calculation of the metrics), and the level of sufficiency of the metric information in the SRS2 after re-work is 87.99% when 8 indicators without considering the number of uses of each indicator and 10 indicators considering the number of uses of each indicator are absent.
The customer of the software of the information and analytical system for the accounting of therapeutic and diagnostic activities provided to the wounded during transportation chose SRS1 with 100%-th level of sufficiency of metric information for further work on a software project.
The developed intelligent system for determining the sufficiency of metric information in the SRS has increased the sufficiency of information by 50.28% for SRS1 and by 12.71% for SRS2. 5
Nowadays the actual task is the calculation of the metrics' values for evaluating and predicting the quality and complexity of software at the early stages of the lifecycle. The conducted analysis of known methods and tools for performing metric analysis showed that most of these methods and tools are focused on the calculation of various metrics based on the analysis of the finished source code. Known methods and tools, which are aimed at early stages of lifecycle, are based on the analysis of the ready values of metrics and don't consider the possibility or impossibility of calculating such metrics on the basis of the available information in the SRS. Then, the actual task is assessing the sufficiency of metric information at the early stages of the software lifecycle, in particular, in the SRS.
In this paper the intelligent system for determining the sufficiency of metric information in the SRS is developed. This IS, based on the natural language processing of SRS, provides: the conclusion about the sufficiency of metric information in SRS, the numerical assessment of the sufficiency level of metric information in the SRS, visualization of missing indicators for metrics calculation.
The developed IS provides the increase in the sufficiency of information by 12.7150.28% for the SRS of the software of the information and analytical system for the accounting of therapeutic and diagnostic activities provided to the wounded during transportation. Generally developed IS provides the increase in the sufficincy of metric information in the SRS to 100% – if it’s necessary (for critical software) or at the request of the customer.
The developed intelligent system for determining the sufficiency of metric information in the SRS can be used in the process of software development for government agencies, military formations and law enforcement agencies, for commercial organizations (for organizations-customers of the software – with the purpose of the assessment of the level of implementation of the initial stages of the lifecycle by the developers and with the purpose of the grounded choice of the SRS with the highest level of sufficiency of information). For organizations-developers, the developed IS can also be used – with the purpose of automation of the process of verifying the sufficiency of information in the SRS.
The limitation of the developed IS is the possibility of assessment of the sufficiency of only metric information in the SRS – as the sufficiency of the indicators in the SRS for determining the 24 selected software metrics, which are available at the design stage. The further efforts of the authors will be directed for elimination of this limitation (in particular, for expanding the set of metrics and, respectively, indicators). 17. Application Intelligence Platform, https://www.castsoftware.com/products/applicationintelligence-platform, last accessed 2019/03/19. 18. ConQAT end of life, https://www.cqse.eu/en/blog/conqat-end-of-life/, last accessed 2019/03/19. 19. GrammaTech CodeSonar: Delivering resiliency for today's IoT devices, https://www.grammatech.com/products/codesonar, last accessed 2019/03/19. 20. Moose is a platform for software and data analysis, http://www.moosetechnology.org/, last accessed 2019/03/19. 21. Automated Software Testing Tools for Creating High Quality Software, https://www.parasoft.com/, last accessed 2019/03/19. 22. SideCI: Automated Code Review for GitHub, https://alternativeto.net/software/sideci/, last accessed 2019/03/19. 23. Sonargraph Product Family, https://www.hello2morrow.com/products/sonargraph, last accessed 2019/03/19. 24. SonarQube: The leading product for continuous code quality, https://www.sonarqube.org/, last accessed 2019/03/19. 25. CppCheck: A tool for static C/C++ code analysis, http://cppcheck.sourceforge.net/, last accessed 2019/03/19. 26. Zakariya, S., Belal, M.: Software quality management measured based code assessments.
International Journal of Computer Science Trends and Technology. 3, 4, 263-268 (2015). 27. Shouman, M., Eldrandaly Kh., Tantawy, A.: Software quality assurance models and expert systems. In: The 9-th International Conference on Production Engineering, Design and Control Proceedings. Alexandria (2009). 28. Agüero, M., Madou, F., Esperón, G., López De Luise, D.: Artificial intelligence for software quality improvement. International Journal of Computer and Information Engineering. 4, 3, 399-404 (2010). 29. Farooq, S. U., Quadri, S., Ahmad, N.: Software measurements and metrics: role in effective software testing. International Journal of Engineering Science and Technology. 3, 1, 671-680 (2011). 30. Gururaj, H. L., Ramesh, B. BornBaby model for software synthesis: A program that can write programs. Data Analytics and Learning: Lecture Notes in Networks and Systems. 43, 403-412 (2019). 31. Ravi Kumar, T., Srinivasa Rao, T.: Software Defects Prediction based on ANN and Fuzzy logic using Software Metrics. International Journal of Applied Engineering Research. 12, 19, 8509-8517 (2017). 32. Azfal, W.: Search-based prediction of software quality: evaluations and comparisons. Blekinge Institute of Technology, Blekinge (2011). 33. Cruickshank, K. J.: A validation metrics framework for safety-critical software-intensive systems. Naval Postgraduate School, Monterey (2009). 34. Michael, J. B., Shing, M.-T., Cruickshank, K. J., Redmond, P. J.: Hazard Analysis and Validation Metrics Framework for System of Systems Software Safety. IEEE Systems Journal. 4, 2, 186–197 (2010). 35. Hovorushchenko, T., Pavlova, O.: Method of Activity of Ontology-Based Intelligent Agent for Evaluating the Initial Stages of the Software Lifecycle. Advances in Intelligent Systems and Computing. 836, 169-178 (2019).