Evaluation Results

RQ1: Does our approach make more relevant recommendations for developers, as compared to existing refactoring techniques?

Median manual evaluations, MC, on the 7 systems.

RQ2: Does our approach significantly reduce the number of relevant refactoring recommendations and the user interaction effort, as compared to existing interactive refactoring approaches?

The median execution time, T, in minutes on the 7 systems.

The median number of required interactions (accept/reject/modify/selection), NI, on the 7 systems.

The median number of recommended refactorings, NR, of the selected solution on the 7 systems.

RQ3: Qualitative Analysis. To what extent the user preferences, interaction and identified region of interests are similar?

Fig. Region of interests at the objective and decision space levels.

Fig. Refactoring types distribution among the solutions selected by the user

Statistical Analysis

Vargha-Delaney: A measure for different metrics between our method (M1) and others.

Label of the methods:

M1 DOIMR (Our approach)
M2=Alizadeh et al. ( V. Alizadeh and M. Kessentini, “Reducing interactive refactoring effort via clustering-based multi-objective search ")
M3=Mkaouer et al. ( V. Alizadeh, M. Kessentini, W. Mkaouer, M. Ocinneide, A. Ouni, and Y. Cai, “An interactive and dynamic search-based approach to software refactoring recommendations,” & M. W. Mkaouer, M. Kessentini, S. Bechikh, K. Deb, and M. O Cinn ´ eide, “Recommendation system for software refactoring ´ using innovization and interactive dynamic optimization,” )
M4=Ouni et al. ( A. Ouni, M. Kessentini, H. Sahraoui, K. Inoue, and K. Deb, “Multicriteria code refactoring using search-based software engineering: An industrial case study,” )

Distribution of the number of required interactions for the 35 participants.

Distribution of the number of refactorings in the selected solutions for the 35 participants

Distribution of the time spent to find a relevant solution for the 35 participants

Google Sites

Report abuse