WP3: Methodology and Quality Assurance Requirements
Type
FUNCTIONAL_SUITABILITY
User Story
As an Operations Engineer/QoS Engineer/Release Manager I want the tool to show me an interpretation of the final decision for the classification of a script as defective.
Requirement
The defect-prediction tool must provide a set of rules that identify defect-prone scripts and an interpretation of the final decision.
Extended Description
The user must have the opportunity to prioritize the actions to solve defects.
Priority
Must have
Affected Tools
DEFECT_PRED_TOOL
Means of Verification
Direct implementation on IDE, feature checklist, case-study
This requirement replaces requirement #4.
The previous requirement stated that the defect prediction tool could provide a defect threat level to architecture elements and predict threat-level defects under certain infrastructure assumptions. This is a typical regression problem that requires to identify the number of bugs in the infrastructure establish a threat-level. It in turns requires an ontology of IaC bugs, which does not exist yet. In addition, the current defect predictor is a classification model based on Decision Tree or Random Forest. Therefore, the requirement has been changed to address its explainability by providing the user with a set of rules that identify defective-prone IaC scripts and the decision path that led to the final prediction. This requirement has also been raised by one of the industrial partners (PRQ) and therefore it’s priority changed from COULD to MUST HAVE.
This requirement replaces requirement #4. The previous requirement stated that the defect prediction tool could provide a defect threat level to architecture elements and predict threat-level defects under certain infrastructure assumptions. This is a typical regression problem that requires to identify the number of bugs in the infrastructure establish a threat-level. It in turns requires an ontology of IaC bugs, which does not exist yet. In addition, the current defect predictor is a classification model based on Decision Tree or Random Forest. Therefore, the requirement has been changed to address its explainability by providing the user with a set of rules that identify defective-prone IaC scripts and the decision path that led to the final prediction. This requirement has also been raised by one of the industrial partners (PRQ) and therefore it’s priority changed from COULD to MUST HAVE.