Supporting material for the paper

Interactive Software Release Planning with Preferences Base


By Altino Dantas, Italo Yeltsin, Allysson Allex Araújo, Jerffeson Souza

Optimization in Software Engineering Group (GOES.UECE) | State University of Ceara - Brazil

1 - Abstract

The release planning is a complex task in the software development process and involves many aspects related to the decision about which requirements should be allocated in each system release. Several search based techniques have been proposed to tackle this problem, but in most cases the human expertise and preferences are not effectively considered. In this context, this work presents an approach which the search is guided according to a Preferences Base supplied by the user. Preliminary empirical results showed the approach is able to find solutions which satisfy the most important user preferences.

Keywords: release planning, interactive genetic algorithm, SBSE.


2 - Instances

Were used two distinct instances composed by real data with 50 and 25 independent requirements obtained from https://sites.google.com/site/mrkarim/data-sets, named as dataset-1 and dataset-2, respectively. The implementation risk of each requirement was randomly assigned. The number of releases of dataset-1 and dataset-2 was fixed to 5 and 8 respectively. The budget for each release was defined as the sum of all requirements costs divided by the number of releases.

Instance Name Number of Files
Clients (M) Requirements (N) Releases (P) Instance List of preferences
Example Instance - - - Download Download
Dataset-1 4 50 5 Download Download
Dataset-2 9 25 8 Download Download

3 - Empirical Studies

Results of SP, SL and Score with μ variation for each instances.

μ dataset-1 dataset-2
SP SL score SP SL score
0 0.40 ±0.03 0.40 ±0.02 25074.8 ±58.33 0.37 ±0.05 0.36 ±0.05 38561.3 ±154.8
0.1 0.54 ±0.01 0.57 ±0.02 24889.8 ±80.55 0.58 ±0.03 0.58 ±0.04 38359.9 ±168.4
0.2 0.62 ±0.02 0.66 ±0.02 24591.2 ±104.23 0.64 ±0.04 0.66 ±0.04 37871.7 ±425.9
0.3 0.65 ±0.02 0.71 ±0.02 24312.2 ±152.30 0.71 ±0.05 0.73 ±0.05 37218.1 ±583.9
0.4 0.74 ±0.02 0.77 ±0.03 23862.6 ±292.98 0.73 ±0.04 0.76 ±0.05 36954.5 ±624.9
0.5 0.75 ±0.03 0.80 ±0.02 23568.0 ±270.29 0.77 ±0.05 0.81 ±0.05 36332.8 ±646.7
0.6 0.77 ±0.02 0.83 ±0.02 23173.3 ±288.83 0.80 ±0.03 0.85 ±0.05 35774.0 ±873.3
0.7 0.80 ±0.03 0.86 ±0.02 22867.4 ±315.07 0.83 ±0.04 0.88 ±0.05 35211.4 ±999.6
0.8 0.81 ±0.02 0.87 ±0.01 22804.4 ±287.04 0.86 ±0.05 0.91 ±0.04 34630.7 ±902.4
0.9 0.82 ±0.02 0.87 ±0.01 22731.9 ±315.73 0.86 ±0.04 0.93 ±0.03 34459.7 ±802.9
1 0.83 ±0.02 0.88 ±0.01 22494.3 ±477.97 0.88 ±0.04 0.94 ±0.04 34052.5 ±674.0

SP: Satisfied Preferences, SL: Sum of importance levels of all satisfied preferences

The symbol means this result is not significantly higher than the previous one, considering the μ variation, (not significantly lower), (significantly higher) and (significantly lower), considerering a 0.05 significance level.