Busari, Saheed Abiola; (2019) Modelling and Analysing Software Requirements and Architecture Decisions under Uncertainty. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Sab_Thesis_Accepted_070219.pdf - Accepted Version Download (5MB) | Preview

Abstract

Early requirements engineering and software architectural decisions are critical to the success of software development projects. However, such decisions are confronted with complexities resulting from uncertainty about the possible impacts of decision choices on objectives; conflicting stakeholder objectives; and a huge space of alternative designs. Quantitative decision modelling is a promising approach to tackling the increasing complexity of requirements and architectural decisions. It allows one to use quantitative techniques, such as stochastic simulation and multi-objective optimisation, to model and analyse the impact of alternative decisions on stakeholders' objectives. Existing requirements and architecture methods that use quantitative decision models are limited by the difficulty of elaborating quantitative decision models and/or lack of integrated tool support for automated decision analysis under uncertainty. This thesis addresses these problems by presenting a novel modelling language and automated decision analysis technique, implemented in a tool called RADAR, intended to facilitate requirements and architecture decisions under uncertainty. RADAR's modelling language has relations to quantitative AND/OR goal models used in requirements engineering and feature models used in software product lines. The language enables modelling requirements and architectural decision problems characterised by (i) single option selection similar to mutually exclusive option selection (XOR-nodes) of feature diagrams; (ii) multiple options selection similar to non-mutually exclusive options selections (OR-nodes) of feature diagrams; and (iii) constraints dependency relationships, e.g., excludes, requires and coupling, between options of decisions. RADAR's analysis technique uses multi-objective simulation optimisation technique in evaluating and shortlisting alternatives that produces the best trade-off between stakeholders' objectives. Additionally, the analysis technique employs information value analysis to estimate the financial value of reducing uncertainty before making a decision. We evaluate RADAR's applicability, usefulness and scalability on a set of real-world systems from different application domains and characterised by design space size between 6 and 2E50. Our evaluation results show that RADAR's modelling language and analysis technique is applicable on a range of real-world requirements and architecture decision problems, and that in few seconds, RADAR can analyse decision problems characterised by large design space using highly performant optimisation method through the use of evolutionary search-based optimisation algorithms.

Download activity - last month

Export as XMLJSONCSV

Download activity - last 12 months

Export as XMLJSONCSV

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item