MOBS Experiment result

This page accompanies a submission to JSS Special Issue on Source Code Analysis and Manipulation, "Evaluting Lexical Approximation of Program Dependence". The page lists all figures arranged in the order listed on paper. It includes the ones omitted from the submission due to space limit.

RQ1 Figure 2. W-ORBS vs. VSM-ORBS vs. LDA-ORBS

Figure 2 shows the comparison between the three slicers, W-ORBS, VSM-ORBS, and LDA-ORBS in a general sense. This page contains plots for all 24 criteria described on the paper.

commons-cli_criterion1 commons-cli_criterion2 commons-cli_criterion3 commons-cli_criterion4 commons-cli_criterion5 commons-csv_criterion1 commons-csv_criterion2 commons-csv_criterion3 guava-escape_criterion1 guava-escape_criterion2 guava-escape_criterion3 guava-escape_criterion4 guava-escape_criterion5 guava-net_criterion1 guava-net_criterion2 guava-net_criterion3 guava-net_criterion4 guava-net_criterion5 printtokens2_criterion1 printtokens_criterion1 replace_criterion1 schedule2_criterion1 schedule_criterion1 totinfo_criterion1

RQ1 Figure 3. VSM-ORBS - Threshold analysis

Figure 3 shows the results of VSM-ORBS with various similarity thresholds(γ). This page contains plots for all 24 criteria described on the paper.

commons-cli_criterion1 commons-cli_criterion2 commons-cli_criterion3 commons-cli_criterion4 commons-cli_criterion5 commons-csv_criterion1 commons-csv_criterion2 commons-csv_criterion3 guava-escape_criterion1 guava-escape_criterion2 guava-escape_criterion3 guava-escape_criterion4 guava-escape_criterion5 guava-net_criterion1 guava-net_criterion2 guava-net_criterion3 guava-net_criterion4 guava-net_criterion5 printtokens2_criterion1 printtokens_criterion1 replace_criterion1 schedule2_criterion1 schedule_criterion1 totinfo_criterion1

RQ1 Figure 4. LDA-ORBS - Threshold analysis

Figure 4 shows the results of LDA-ORBS with various similarity thresholds(γ). This page contains plots for all 24 criteria described on the paper. We chose the topic counts(n) as 500 for Java and 200 for C.

commons-cli_criterion1 commons-cli_criterion2 commons-cli_criterion3 commons-cli_criterion4 commons-cli_criterion5 commons-csv_criterion1 commons-csv_criterion2 commons-csv_criterion3 guava-escape_criterion1 guava-escape_criterion2 guava-escape_criterion3 guava-escape_criterion4 guava-escape_criterion5 guava-net_criterion1 guava-net_criterion2 guava-net_criterion3 guava-net_criterion4 guava-net_criterion5 printtokens2_criterion1 printtokens_criterion1 replace_criterion1 schedule2_criterion1 schedule_criterion1 totinfo_criterion1

RQ1 Figure 5. LDA-ORBS - Impact of topic count

Figure 5 shows the results of LDA-ORBS with various topic counts(n). This page contains plots for all 24 criteria described on the paper. We chose the threshold(γ) as 0.9.

commons-cli_criterion1 commons-cli_criterion2 commons-cli_criterion3 commons-cli_criterion4 commons-cli_criterion5 commons-csv_criterion1 commons-csv_criterion2 commons-csv_criterion3 guava-escape_criterion1 guava-escape_criterion2 guava-escape_criterion3 guava-escape_criterion4 guava-escape_criterion5 guava-net_criterion1 guava-net_criterion2 guava-net_criterion3 guava-net_criterion4 guava-net_criterion5 printtokens2_criterion1 printtokens_criterion1 replace_criterion1 schedule2_criterion1 schedule_criterion1 totinfo_criterion1

RQ1 Figure 6. VSM-ORBS-M and LDA-ORBS-M

Figure 6 shows the results of VSM-ORBS-M and LDA-ORBS-M. This page contains 48 plots, one from each VSM-ORBS-M and LDA-ORBS-M for all 24 criteria described on the paper.

vsm-inc_commons-cli_criterion1 vsm-inc_commons-cli_criterion2 vsm-inc_commons-cli_criterion3 vsm-inc_commons-cli_criterion4 vsm-inc_commons-cli_criterion5 vsm-inc_commons-csv_criterion1 vsm-inc_commons-csv_criterion2 vsm-inc_commons-csv_criterion3 vsm-inc_guava-escape_criterion1 vsm-inc_guava-escape_criterion2 vsm-inc_guava-escape_criterion3 vsm-inc_guava-escape_criterion4 vsm-inc_guava-escape_criterion5 vsm-inc_guava-net_criterion1 vsm-inc_guava-net_criterion2 vsm-inc_guava-net_criterion3 vsm-inc_guava-net_criterion4 vsm-inc_guava-net_criterion5 vsm-inc_printtokens2_criterion1 vsm-inc_printtokens_criterion1 vsm-inc_replace_criterion1 vsm-inc_schedule2_criterion1 vsm-inc_schedule_criterion1 vsm-inc_totinfo_criterion1 lda-inc_commons-cli_criterion1 lda-inc_commons-cli_criterion2 lda-inc_commons-cli_criterion3 lda-inc_commons-cli_criterion4 lda-inc_commons-cli_criterion5 lda-inc_commons-csv_criterion1 lda-inc_commons-csv_criterion2 lda-inc_commons-csv_criterion3 lda-inc_guava-escape_criterion1 lda-inc_guava-escape_criterion2 lda-inc_guava-escape_criterion3 lda-inc_guava-escape_criterion4 lda-inc_guava-escape_criterion5 lda-inc_guava-net_criterion1 lda-inc_guava-net_criterion2 lda-inc_guava-net_criterion3 lda-inc_guava-net_criterion4 lda-inc_guava-net_criterion5 lda-inc_printtokens2_criterion1 lda-inc_printtokens_criterion1 lda-inc_replace_criterion1 lda-inc_schedule2_criterion1 lda-inc_schedule_criterion1 lda-inc_totinfo_criterion1

RQ2 Figure 7. VSM-ORBS and LDA-ORBS followed by W-ORBS

Figure 7 compares the results of applying VSM-ORBS or LDA-ORBS followed by W-ORBS to those attained by applying W-ORBS alone.

vsm-inc_commons-cli_criterion1

RQ3 Figure 8. Vargha-Delaney Â12 on SPD between MOBS selection strategies

Figure 8 shows the result of Vargha-Delaney Â12 statistic for the SPD values of the four different operator selection strategies used by MOBS: ROS, APP, AFF, and UNI.

commons-cli_criterion1

RQ4 Figure 9. Time & deleted lines for W-ORBS and MOBS

Figure 9 compares the wall colock execution time and the number of deleted lines after each iteration between W-ORBS and MOBS of four selection strategies. The result of ten trials of each strategies is represented as a boxplot. This page contains plots for all 24 criteria described on the paper.

commons-cli_criterion1 commons-cli_criterion2 commons-cli_criterion3 commons-cli_criterion4 commons-cli_criterion5 commons-csv_criterion1 commons-csv_criterion2 commons-csv_criterion3 guava-escape_criterion1 guava-escape_criterion2 guava-escape_criterion3 guava-escape_criterion4 guava-escape_criterion5 guava-net_criterion1 guava-net_criterion2 guava-net_criterion3 guava-net_criterion4 guava-net_criterion5 printtokens2_criterion1 printtokens_criterion1 replace_criterion1 schedule2_criterion1 schedule_criterion1 totinfo_criterion1

RQ5 Figure 10. S.Jaccard similarity between ROS-MOBS and W-ORBS

Figure 10 shows S.Jaccard similarity between ROS-MOBS and W-ORBS. This page contains plots for all 24 criteria described on the paper.

commons-cli_criterion1 commons-cli_criterion2 commons-cli_criterion3 commons-cli_criterion4 commons-cli_criterion5 commons-csv_criterion1 commons-csv_criterion2 commons-csv_criterion3 guava-escape_criterion1 guava-escape_criterion2 guava-escape_criterion3 guava-escape_criterion4 guava-escape_criterion5 guava-net_criterion1 guava-net_criterion2 guava-net_criterion3 guava-net_criterion4 guava-net_criterion5 printtokens2_criterion1 printtokens_criterion1 replace_criterion1 schedule2_criterion1 schedule_criterion1 totinfo_criterion1

RQ5 Figure 11 & 12. Deleted characterized lines and non-stop words

First three plots on this page shows the number of lines that remain only on the slice from ROS-MOBS but have been removed on the slice from W-ORBS. The last plot shows an average number of non-stop word tokens in a single deleted line.

commons-cli_criterion1 commons-cli_criterion2 commons-cli_criterion3 commons-cli_criterion4

RQ6 Figure 13. Scalability - MOBS on Misaka

Figure 13 shows the result comparing W-ORBS and ROS-MOBS by showing the ratio between two. The red box plot on the left represents how many times ROS-MOBS run faster compare to W-ORBS, and the blue box plot on the right represents the ratio of the number of deleted lines by ROS-MOBS to the number of deleted lines by W-ORBS.

commons-cli_criterion1