Zobraziť v EDS

Hybrid semi-supervised SOM based clustered approach with genetic algorithm for software fault classification.

Uložené v:
Podrobná bibliografia
Názov: Hybrid semi-supervised SOM based clustered approach with genetic algorithm for software fault classification.
Autori: Aarti1 (AUTHOR) aarti.25209@lpu.co.in, Rajput, Pushpendra Kumar2 (AUTHOR) pushpendra@ddn.upes.ac.in, Khare, Ankit3 (AUTHOR) ankitkhare@srhu.edu.in
Zdroj: AIP Conference Proceedings. 2023, Vol. 2724 Issue 1, p1-8. 8p.
Predmety: *GENETIC software, *GENETIC algorithms, *SELF-organizing maps, *COMPUTER software industry, *SUBSET selection, *FEATURE selection
Abstrakt: The delivery of defect-free products is always being a challenge in the software industry. Limitation of testing criteria is reasoned as important aspects that lead to the existence of faults/bugs in the developed system. However, fault and effort prediction is a futuristic event in any software development-planning phase. Nevertheless, to save time, effort and budget forecasting faults and effort become critical aspects of software development. It has been proven that unsupervised and semi-supervised classification techniques produce more accurate results in the lack of availability of past information. To reduce the manual intervention of experts for identifying modules, authors propose an automatic software tool with a semi-supervised feature based on a self-organizing map to detect labels using reduced map size. Three different scenarios, which integrate proposed clustering with regression-based classification, are the main contribution of the study. The fusion of clustering and regression improves the capability of the prediction model in the presence of heterogeneous data. The use of feature subset selection is also considered with an experimental comparison. The combination of feature selection with the proposed technique provides more flexibility to choose a significant amount of attributes. [ABSTRACT FROM AUTHOR]
Databáza: Academic Search Index
Popis
Abstrakt:The delivery of defect-free products is always being a challenge in the software industry. Limitation of testing criteria is reasoned as important aspects that lead to the existence of faults/bugs in the developed system. However, fault and effort prediction is a futuristic event in any software development-planning phase. Nevertheless, to save time, effort and budget forecasting faults and effort become critical aspects of software development. It has been proven that unsupervised and semi-supervised classification techniques produce more accurate results in the lack of availability of past information. To reduce the manual intervention of experts for identifying modules, authors propose an automatic software tool with a semi-supervised feature based on a self-organizing map to detect labels using reduced map size. Three different scenarios, which integrate proposed clustering with regression-based classification, are the main contribution of the study. The fusion of clustering and regression improves the capability of the prediction model in the presence of heterogeneous data. The use of feature subset selection is also considered with an experimental comparison. The combination of feature selection with the proposed technique provides more flexibility to choose a significant amount of attributes. [ABSTRACT FROM AUTHOR]
ISSN:0094243X
DOI:10.1063/5.0141332