D.3.3 ALGORITHMS FOR INCREMENTAL ... - SecureChange
D.3.3 ALGORITHMS FOR INCREMENTAL ... - SecureChange
D.3.3 ALGORITHMS FOR INCREMENTAL ... - SecureChange
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Evolution<br />
Elicitation<br />
Probability<br />
Estimation<br />
Reasoning<br />
Decision<br />
Fig. 2. Requirements Engineering process for Change/Evolution Management<br />
The elicitation of evolution rules is the first step of the change management illustrated<br />
in Fig. 2. The process consists of four major steps, namely Evolution elicitation,<br />
Probability Estimation, Reasoning, and Decision [14].<br />
STEP 1 Evolution Elicitation. The goal of this key step is to extract evolution rules<br />
from a description in natural language of the possible requirements evolutions.<br />
In order to do that, first, statements about changes are identified. Then, changes<br />
are clustered into mutually exclusive groups. Then, each group will represent an<br />
After i requirement model.<br />
STEP 2 Probability Estimation. Based on the evolution rules identified in the previous<br />
step, this step relies on methods like Analytic Hierarchy Process (AHP) to identify<br />
the evolution probability for each possible evolution After i of an observable evolution<br />
rule. The probabilities are then validated with stakeholders(and/or domain<br />
experts) using a game-theoretic approach [15] to ensure that they are meaningful.<br />
STEP 3 Reasoning. Based on the representation of an evolving requirement model as<br />
a set of controllable and observable rules, it is possible to compute two quantitative<br />
metrics called maximal belief and residual risk that intuitively measure the usefulness<br />
of a model element (or a set of elements) after evolution. In fact, the maximal<br />
belief tells whether a design alternative is useful after evolution, while residual risk<br />
quantifies if a design alternative is no longer useful.<br />
STEP 4 Decision. Decision makers take the analysis’ outcome to choose the optimal<br />
design for the next development phases. Depending on the kind of analysis, different<br />
“optimal” criterion can be applied. For instance, a selection criterion for the<br />
analysis is: “higher max belief and lower residual risk”.<br />
4 Research Methodology<br />
Many different assessment methodologies can be used for validation with domain experts.<br />
For this study, we have mainly used qualitative techniques. Qualitative research<br />
consists of an application of various methods of collecting information, mainly through<br />
focus groups and interviews. Interviews are commonplace techniques where domain<br />
experts are asked questions by an interviewer in order to gain domain knowledge and<br />
it is the most widely used method of finding out what users want [8]. Focus groups<br />
bring together a cross-section of stakeholders in an informal discussion group format.<br />
The user study has involved twelve participants having different background and roles<br />
as summarized in Table 2. P1, P2, P3 and P4 played the role of requirement analyst<br />
and observer. Requirement analysts were responsible for the organization of the user<br />
study and for the analysis of the results. The observers were responsible to record the<br />
meetings and take notes during the execution of the study. The other participants were<br />
ATM domain experts who participated to the semi-structured validation of the proposed<br />
approach to model and reason on requirements evolution.<br />
5