Academia.eduAcademia.edu

Figure 3 – uploaded by Shorouq AL-eidi

See full PDFdownloadDownload figure
From this case study, we captured the software complexity’s influence on the readabil- ity of code. It is intuitive as the readability reflects many attributes, including code structure and density. Thus, there must be an overlap between those attributes. The experiment confirmed the HO hypothesis, which claimed that software complexity had an influence on the readability of code, and proved the invalidity of hypothesis H1, which claimed that software complexity had no influence on the readability of code. Moreover, the results indicated that software complexity affected code readability with 90.01% accuracy using a decision tree classifier. Therefore, it was considered a strong influence. Figure 3 presents the first three levels of the tree model. We used only the first three levels due to the large tree size; these levels showed the most effective complexity metrics in this relation, which were Halstead volume and OSavg. Figure 3. First three levels of the model for case study 2.

Figure 3 From this case study, we captured the software complexity’s influence on the readabil- ity of code. It is intuitive as the readability reflects many attributes, including code structure and density. Thus, there must be an overlap between those attributes. The experiment confirmed the HO hypothesis, which claimed that software complexity had an influence on the readability of code, and proved the invalidity of hypothesis H1, which claimed that software complexity had no influence on the readability of code. Moreover, the results indicated that software complexity affected code readability with 90.01% accuracy using a decision tree classifier. Therefore, it was considered a strong influence. Figure 3 presents the first three levels of the tree model. We used only the first three levels due to the large tree size; these levels showed the most effective complexity metrics in this relation, which were Halstead volume and OSavg. Figure 3. First three levels of the model for case study 2.

Related Figures (12)

Figure 1. Research methodology flowchart. SEDER Moca te TRAE SER USS LO PSRD ORC EERO GHEE See PERE SSE USE NES GER, OSES SSE CRUSE REIS SSRIS IORSRSEL IGS THEE SSSEIRISE IP SSRIS ERESE SASS SSE REESE OE SOTER, Figure 1 presents the research methodology flowchart. First, we downloaded Java source code files by utilizing Eclipse, which is a well-known software development envi- ronment and open-source software. We then extracted various software metrics from these source code files, which were related to software readability and complexity. After that, we built our data set. In the next step, we applied some statistical techniques proposed in this paper to combine the readability metrics or complexity metrics into a single variable for further analysis. Then, we used analysis methods and data mining techniques, including machine learning algorithms, to test several research hypotheses. Finally, we discussed the results and presented our contribution. The later sections briefly explain the processes for the research work.
Figure 1. Research methodology flowchart. SEDER Moca te TRAE SER USS LO PSRD ORC EERO GHEE See PERE SSE USE NES GER, OSES SSE CRUSE REIS SSRIS IORSRSEL IGS THEE SSSEIRISE IP SSRIS ERESE SASS SSE REESE OE SOTER, Figure 1 presents the research methodology flowchart. First, we downloaded Java source code files by utilizing Eclipse, which is a well-known software development envi- ronment and open-source software. We then extracted various software metrics from these source code files, which were related to software readability and complexity. After that, we built our data set. In the next step, we applied some statistical techniques proposed in this paper to combine the readability metrics or complexity metrics into a single variable for further analysis. Then, we used analysis methods and data mining techniques, including machine learning algorithms, to test several research hypotheses. Finally, we discussed the results and presented our contribution. The later sections briefly explain the processes for the research work.
Table 1. Clustering technique results.
Table 1. Clustering technique results.
Table 3. Case study 1 results for summation division approach using percentage split. Table 4. Case study 1 results for summation division approach using cross-validation.
Table 3. Case study 1 results for summation division approach using percentage split. Table 4. Case study 1 results for summation division approach using cross-validation.
Figure 2. First three levels of the model for case study 1. From this case study, we found that code readability had a great influence and impact on software complexity, which addressed the essential role that code readability plays on the maintainability process and on overall software quality. The experimental results confirmed the HO hypothesis which claimed that code readability had an influence on software complexity and proved the invalidity of hypothesis H1 which claimed that code readability had no influence on software complexity. Moreover, the results indicated that code readability affected software complexity with a 90.15% accuracy using a decision tree classifier. Therefore, it was considered as a strong influence. Figure 2 presents the first three levels of the tree model. Due to the tree’s large size, we only show the first three levels; these levels show the most effective readability features in this relation which are: max character occurrences, average arithmetic, average parenthesis, max indents, and max indent (preceding white space).
Figure 2. First three levels of the model for case study 1. From this case study, we found that code readability had a great influence and impact on software complexity, which addressed the essential role that code readability plays on the maintainability process and on overall software quality. The experimental results confirmed the HO hypothesis which claimed that code readability had an influence on software complexity and proved the invalidity of hypothesis H1 which claimed that code readability had no influence on software complexity. Moreover, the results indicated that code readability affected software complexity with a 90.15% accuracy using a decision tree classifier. Therefore, it was considered as a strong influence. Figure 2 presents the first three levels of the tree model. Due to the tree’s large size, we only show the first three levels; these levels show the most effective readability features in this relation which are: max character occurrences, average arithmetic, average parenthesis, max indents, and max indent (preceding white space).
Table 9. Case study 2 results for PCA dimension reduction approach using percentage split. Table 10. Case study 2 results for PCA dimension reduction approach using cross-validation
Table 9. Case study 2 results for PCA dimension reduction approach using percentage split. Table 10. Case study 2 results for PCA dimension reduction approach using cross-validation
Table 11. LOC and readability metrics relation. 4.4.1. Summation Division Results
Table 11. LOC and readability metrics relation. 4.4.1. Summation Division Results
Table 12. LOC and readability metrics results for summation division approach using percentage split. Table 13. LOC and readability metrics results for summation division approach using cross- validation.
Table 12. LOC and readability metrics results for summation division approach using percentage split. Table 13. LOC and readability metrics results for summation division approach using cross- validation.
4.4.2. PCA Dimension Reduction Results
4.4.2. PCA Dimension Reduction Results
Table 14. LOC and readability metrics results for PCA dimension reduction approach using percent- age split. Table 15. LOC and readability metrics results for PCA dimension reduction approach using cross- validation.
Table 14. LOC and readability metrics results for PCA dimension reduction approach using percent- age split. Table 15. LOC and readability metrics results for PCA dimension reduction approach using cross- validation.
Based on our experiments and results, we found that readability features were depen- dent on the code size, where the code size had an influence on code readability features. The accuracy results illustrated the strong influence and relationship that the readability features have on the code size, which confirmed the HO hypothesis and consequently proved the invalidity of hypothesis H1. Therefore, we illustrated in this case study that readability was affected by the attribute size directly and indirectly.
Based on our experiments and results, we found that readability features were depen- dent on the code size, where the code size had an influence on code readability features. The accuracy results illustrated the strong influence and relationship that the readability features have on the code size, which confirmed the HO hypothesis and consequently proved the invalidity of hypothesis H1. Therefore, we illustrated in this case study that readability was affected by the attribute size directly and indirectly.
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved